Revadim is a multi-layered late Acheulian site located on the southern coastal plain of Israel (Figure 1A). Paleomagnetism results for the sequence exposed in the quarry showed normal polarity, indicating that the sequence is younger than 780,000 BP. Uranium-thorium analysis of carbonate covering flint items dated it to between 500,000 and 300,000 BP. During four excavation seasons, four main areas were excavated (A–D) as well as two trenches which stratigraphically connect Areas C and B. Altogether, an area of 250 m² was excavated, 170 m² were exposed in Areas A–D, and ca. 80 m² in Trenches 12 and 23. The analysis of the flint and faunal assemblages indicates a late Acheulian assignment of all layers of the site (see details in Gvirtzman et al., 1999; Marder et al., 1999; Marder et al., 2011; Rabinovich et al., 2012; Solodenko et al., 2015; Agam and Barkai, 2018a; Zupancich et al., 2018; Rosenberg-Yefet and Barkai, 2019; Rosenberg-Yefet et al., 2021). The rich faunal assemblage includes Palaeoloxodon antiquus, Bos primigenius, Gazella gazelle, Capra cf. aegagrus, and Dama cf. mesopotamica (Gvirtzman et al., 1999; Rabinovich et al., 2012).

Previous analysis of the Revadim site revealed a varied core technology that includes three main reduction sequences: 1) The production of large and medium flakes from single, double or multiple striking platform cores, 2) the production of predetermined items from prepared cores and discoid cores, and 3) the production of very small flakes by recycling of existing items as cores on flakes (Malinsky-Buller et al., 2011; Agam et al., 2015; Solodenko et al., 2015; Rosenberg-Yefet, 2016; Agam and Barkai, 2018a; Venditti et al., 2019a; Venditti et al., 2019b).

The blanks produced by these methods were sometimes further modified in different ways and constitute the largest group of tools, termed flake-tools, alongside a wide variety of bifaces and other tool types (e.g., handaxes and chopping tools) (Marder et al., 2006; Solodenko, 2010; Malinsky-Buller et al., 2011; Agam et al., 2015; Solodenko et al., 2015; Malinsky-Buller, 2016; Cohen, 2018; Rosenberg-Yefet and Barkai, 2019).

The sample chosen for this study includes prepared cores and items detached from prepared cores, originating from four different contexts of the site, representing the four lithic assemblages studied thus far. These are layers B2 and C5 (Figure 1B), which are the oldest contexts of the site and assumed to be rather chrono-stratigraphically contemporaneous, and layers C3 and B1, which are stratigraphically younger. All four assemblages have been initially sorted and analyzed (Supplementary Table S1).

Core trimming elements (CTEs), which are the products of shaping and maintaining prepared cores and originated from the studied layers (C3 and C5), were also included. These items were separated from the rest of the CTE assemblage during preliminary classification.

Jaljulia is a late Acheulian site located on the central part of the coastal plain of Israel (Figure 1A), some 6 km south of the late Acheulian site of Eyal and 6 km north-west of the Acheulo-Yabrudian site of Qesem Cave (Ronen and Winter 1997; Gopher et al., 2005). Approximately 80 m² of late Acheulian deposits were excavated during 2016–2017 in the framework of a salvage excavation by the Israel Antiquities Authority in collaboration with the Department of Archaeology at Tel-Aviv University (Shemer et al., 2018). Lithic assemblages are rich (Figure 1C) but faunal preservation is poor. Lithic analysis is currently in process, and therefore only preliminary observations are currently available (Supplementary Table S2). The assemblages are mostly dominated by flake production and flake-tools, although all include handaxes. Core technology varies and includes three main reduction sequences, very much similar to those described above for Revadim. The lithic assemblage chosen for this study, currently at the final stages of lithic analysis, was retrieved from an area of 6 m² (area B, Figure 1D).

A total of 72 prepared cores from Revadim (out of a total of 1407 cores from the four assemblages, 5.1% of all cores) and a total of 105 prepared cores, including 13 core fragments, (out of a total of 602 cores, 17.4% of all cores) from Jaljulia were included in the current study, with a further subdivision into proto-Levallois cores, discoid cores, and prepared cores (general) (Supplementary Figure S1) for those that could not be assigned to the Levallois/discoid categories. See Supplementary Table S3 for details.

Characterizing prepared cores is difficult due to the variety of definitions used in the literature and the lack of uniformity. While some scholars use the term “prepared cores” to describe cores similar to Levallois cores (e.g., Wilkins et al., 2010; Leader et al., 2018), we use the term more generally to include proto-Levallois cores but also other core methods. Our definition is in the spirit of Debénath and Dibble (1994:23), who defined prepared cores as “a number of technologies…in which the core was intentionally shaped or prepared in such a way as to predetermine the shapes of flakes taken from it”. All cores under that definition are fully or partially centripetal in their design, in contrast to one, two or multi-platform cores. Therefore, all cores are characterized by two surfaces and a plane of intersection, although their exact role in the reduction sequence varies between sub-methods, as does the presence or absence of hierarchy between them.

Cores originating from pre-Middle Paleolithic contexts but bearing conceptual resemblance to Levallois cores are defined in very different ways by different scholars. We follow here the definition proposed by Picin (2017), whereby proto-Levallois cores are cores in which the volume is divided into two hierarchical surfaces, a striking platform and a flaking surface, with a plane of intersection separating the two surfaces. The plane of intersection is delineated by a partial or complete bifacial ridge. Striking platforms are usually roughly prepared, and lateral and distal convexities of the flaking surface are roughly configured by either débordants flakes or by preparational flaking of the core circumference (Picin 2017). In this study we did not define cores as fully fledged Levallois, though some of the cores do represent a more developed stage of the technology’s adoption and some do meet all Boëda’s criteria. The main technological difference concentrates on less strict application of the criteria. For instance, core circumference is not always prepared fully and lateral and distal convexities of the flaking surface are sometimes only roughly designed in Proto-Levallois cores.

The discoid method was the most common core technology within the prepared core category from the Lower Paleolithic onwards, although it was recognized and defined as Mousterian debitage at earlier stages of research (González-Molina et al., 2020).

On the basis of definitions by both Boëda in the mid-1990s and Terradas (2003), the criteria for identifying discoid cores were recently summarized by González-Molina et al. (2020): 1) the volume of the core is conceived in two asymmetric, secant and convex surfaces that delimit an intersection plane. 2) There is no hierarchization between the two surfaces of the core: one is conceived as a flaking surface and the other as striking platform, but with the possibility of role inversion during one operational sequence. 3) The exploitation or debitage surface is designed with a peripheral convexity that controls the knapping of each extraction. 4) The flaking axis of predetermined removals is perpendicular to the striking platform employed to obtain it. 5) The striking surface is oriented so that the intersection between both surfaces is perpendicular to the edge of the core. 6) The technique is direct percussion using a hard hammer.

Because the discoid method shares four of the six criteria described by Boëda (1993) (detailed in the introduction), further clarification is required to distinguish the two methods. The discovery and analysis of variety of discoid cores, both chronologically and geographically distributed, as part of a collection of studies by Peresani (2003), have helped researchers to resolve some of the ambiguities by distinguishing between the sensu stricto and sensu lato definitions of discoid cores (Picin and Vaquero, 2016). The term discoid sensu lato describes better cores that do exhibit hierarchy between the two surfaces, whereas sensu stricto describes cores that differ from the Levallois in the lack of hierarchy between their surfaces and in the direction of production, which is secant to the plane of intersection of the two surfaces (Picin and Vaquero, 2016). For a recent study on the distinctions between these two methods, see González-Molina et al. (2020)

Our definition of discoid cores follows the guidelines set out by Terradas, who sees the discoid flaking method as consisting of “the core exploitation in order to obtain various numbers of flakes by centripetal, recurrent and usually bifacial organization of removals. This exploitation proceeds by maintaining a strong stability of the volumetric concept of the core, which requires little specific preparation of its striking platforms and flaking surfaces. The core resulting from this sort of exploitation has an oval shape and a biconvex asymmetric section” (Terradas, 2003). These cores comply with the sensu lato definition, since the term refers to cores that do exhibit hierarchy between the two surfaces, which in our case are also switchable.

We analyzed the cores by weight (light, 1–49 g, medium, 50–99 g, and heavy, above 100 g), flint quality (homogeneous, fairly homogeneous, and heterogeneous), and texture (fine-grained, medium-grained, and coarse). Determining flint quality and its effect on toolmaking is challenging. While microscopic material evaluations are gaining a foothold, this is out of the scope of the current research. Therefore, here we determined flint quality and texture as visible to the naked eye. Stone quality, although a subjective criterion, was determined in this study by the homogeneity of the material and by the presence or absence of disturbances in the flint. Homogeneous flint refers to flint with no disturbances or cracks. Fairly homogeneous flint will show one or two disturbances or cracks, and heterogeneous flint will show several. A detailed analysis of prepared core flint types from Jaljulia appears in Agam et al. (submitted).

Blank types include nodules, pebbles, old cores\handaxes, flakes, and undetermined. Handaxes that were recycled as cores while undergoing some modifications were included here, whereas handaxes with a preferential flake scar having minimal preparations were not included in the core category and are not part of the current study (see Rosenberg-Yefet et al., 2021 for a detailed description of these items). The presence of cortex on the striking platform and on the production surface was also noted, as was the presence of flaked patinated surfaces as an indication for the use of old flaked items as cores. Scar characteristics include the number of shaping scars on the production surface, the number of shaping scars on the striking platform, and the number of unsuccessful removal scars on each surface. Unsuccessful removals were determined by scars with hinge or step termination, scars indicating an angle which damaged core morphology, or scars that differ from the others in shape, depth or thickness. Scars that cross the midline of the production surface were also indicated. The minimum and maximum angles between the two platforms of the cores were measured. Core convexities were estimated (% of the core circumference). The presence of éclats débordants scars was noted, as were indications of striking platform preparation following the detachment of predetermined blanks. (By this we mean preparations that appear to have been made in the process of producing another item but the core was abandoned before that item was produced.) The core exploitation method (preferential, recurrent centripetal, recurrent unidirectional, or recurrent bidirectional) was also indicated. The detailed criteria are listed in Table 1, Table 2, and Table 3.

CTEs were categorized as either CTEs removed from prepared cores or éclats débordants, with the latter category further subdivided into éclats débordants, éclats débordants a dos limité, and éclats outrepassés. CTES removed from prepared cores are items for which part of the intersecting ridge between the two platforms was removed, most probably in an effort to maintain core convexities throughout the reduction sequence. These items do not fully conform to the definition of the specific CTEs detailed below.

The éclats débordants items are defined as follows:

Éclats débordants (Supplementary Figure S2A): The terminology is drawn from similar items used in the application of the Middle Paleolithic Levallois method (Beyries and Boëda, 1983; Geneste, 1985:230; Meignen, 1995:280). It describes items removed from the lateral edge of a core during the shaping and maintenance of the production surface, mostly between production phases. Part of the bifacially shaped plane of intersection that separated the production surface from the striking platform was also removed (Boëda, 1993). Such removals took place during relatively advanced stages of the Levallois reduction sequence (Hovers, 2009:82).

Éclats débordants a dos limité (Supplementary Figure S2B), also termed pseudo-Levallois flakes: These items were produced during advanced stages of Levallois core reduction (Boëda et al., 1990:68; Hovers, 2009:82) and were removed from the core in order to correct irregularities in the morphology of the production surface, thus contributing to the maintenance of lateral convexities during the different phases of the Levallois reduction sequence (Boëda et al., 1990).

Éclats outrepassés (Supplementary Figure S2C). “These were detached only when intensive use of both the striking platform and the flaking surface of the core had distorted the geometry of the flaking surface. Rejuvenation of the lateral and (mainly) distal convexities required removal of the exploited surface and extensive rearrangement of the flaking surface. Such items are expected to show a large number of dorsal face scars and a complex dorsal face scar pattern” (Hovers, 2009:82).

Similarities between CTEs removed from prepared cores and items removed from handaxes during their different manufacturing stages are also noteworthy. These similarities, mainly in dorsal scar pattern and the removal of parts of the bifacial ridge, make it difficult to distinguish between these two categories (Copeland, 1995; Goren-Inbar et al., 2018).

CTEs were analyzed according to the following parameters: metric measurements (maximum length and width), scar pattern (centripetal, unidirectional, or bidirectional) and number of scars, platform preparation type (plain, punctiform, faceted, dihedral, linear, or undetermined), interior and exterior platform angle (IPA and EPA, respectively), ridge angle, cross section type (obtuse triangle, isosceles trapezoid, parallelogram, equilateral triangle, right triangle), and end termination (feather, overshot, hinge, retouched, or unclear).

For the platform angles, we followed Dibble and Whittaker (1981: 286), who define IPA as “the angle formed between the platform surface and a line through the point of percussion to the base of the bulb” and EPA as “measured at the intersection of the platform surface and the exterior surface of the core” (Supplementary Figure S3).

Blanks produced from prepared cores are predetermined items that were detached from prepared cores. It seems that these blanks were not produced for the purpose of shaping or maintaining the core; we thus view them as the desired end-products of the knapping process. They usually exhibit a larger IPA than the regular flakes, and their striking platforms are usually shaped, faceted, or dihedral. Their dorsal scar pattern varies, but it is either centripetal or unidirectional. If unidirectional, it differs from the scar pattern that we see in these sites for the regular flakes. These items were retrieved from all other categories (shaped and unshaped items) following the identification of particular characteristics (shaped striking platform, scar pattern, curvature, and so forth). Items detached from prepared cores were analyzed according to the following parameters: metric measurements (including thicknesses in the bulb area, middle and end of the flake), scar pattern and number of scars, platform preparation type, IPA and EPA, and end termination.

The major attributes and metrics of the CTE assemblages from Revadim and Jaljulia are presented in Table 5 and Supplementary Table S5.

Éclats débordants (Revadim: n = 11; Jaljulia: n = 20): This group was further divided into éclats débordants (Figures 8A–D; Figures 9A,B), éclats débordants a dos limité (Figure 9D), and outrepassés (Figure 8E; Figure 9C). Due to the small sample size of each group, the analysis below will be presented for all sub-types together. Most items are characterized by a centripetal scar pattern on the dorsal face while the minority are unidirectional or bidirectional, with an average of seven scars on the dorsal face for Revadim and 5.3 for Jaljulia. The items have quite similar average EPA (Revadim: 82°; Jaljulia: 83.1°), reflecting the angle between the two platforms of the core from which they were removed, and an average IPA of 112° for Revadim and 105.6° for Jaljulia, indicating the angle of removal of the items. Variability in striking platform preparation was also noted, as some are plain, punctiform, faceted, dihedral, linear, or undetermined. At Revadim, only one item was removed from the right lateral part of the core, while the rest were removed from the left lateral part of the core (Figure 10A). At Jaljulia, most items were removed from the right lateral part of the core, while only a few were removed from the left lateral part of the core (Figure 10B). The average angle of the ridge removed from the core is quite similar for both sites (Revadim: 62.7°; Jaljulia: 67.7°). The EPA and ridge angle both mirror the original core circumference at different locations on the core. The difference between the EPA and the ridge angle is relatively large (about 20°), which might indicate a calculated decision on the part of the knapper as to where to strike the blow, even for centripetal cores that seem at first glance to be rounded and symmetrical. The cross sections of most items are obtuse triangles, while the rest are isosceles trapezoids, right triangles, parallelograms, or undetermined. The end terminations of the items vary between overshot (Figure 8E; Figure 9C), feather (Figures 8A,C; Figure 9A), hinge (Figures 8B,D), and broken\undetermined. (See Figure 9D).

Prepared core CTEs (Revadim: n = 10; Jaljulia: n = 40): Most items bear a centripetal scar pattern on the dorsal face while the minority are unidirectional or bidirectional, with an average of six scars on the dorsal face for Revadim and 5.6 for Jaljulia. The items have quite similar average EPA (Revadim: 73°; Jaljulia: 78.7°), reflecting the angle between the two platforms of the core from which they were removed, and an average IPA of 111° for Revadim and 110.1° for Jaljulia, indicating the angle of removal of the items. Variability in striking platform preparation was also noted, as some are plain, punctiform, faceted, dihedral, linear (or undetermined. At Revadim, half of the items were removed from the left lateral part of the core and 40% (n = 4) from the right lateral part of the core (Figure 10A); the remaining 10% were removed from the proximal end. Similarly, at Jaljulia, about half of the items were removed from the left lateral part of the core, and the other half from the right lateral part of the core (Figure 10B); the remaining 5% were removed from the proximal end. The average angle of the ridge removed from the core is quite similar for both sites (Revadim: 70°; Jaljulia: 66.1°). Here, too, the large difference between the EPA and the ridge angle might indicate a deliberate decision as to where to strike the core, as explained above. Most items have an obtuse triangular cross section, while the rest consist of a few different cross section types (Table 5). The end terminations of the items vary, but the most common are overshot, and hinge (Table 5).

The frequency of blanks produced from prepared cores differs significantly for the two sites (Jaljulia, n = 99; Revadim, n = 8). Since the number of items from Revadim is very small, quantitative analysis is not possible and thus the data will be presented separately for each site.

We analyzed the techno-typological characteristics of prepared cores and blanks originating from four layers of Revadim and one excavation area from Jaljulia, including all observable aspects of the core reduction sequence. Three sub-types of prepared cores were defined at Revadim and Jaljulia: prepared cores (general), discoid cores, and proto-Levallois cores. Proto-Levallois and prepared cores (general) adhere to concepts characteristic of the Levallois method. We thus believe that the production of predetermined blanks in the late Acheulian should be viewed as a precursor of the fully-fledged Levallois method of the Middle Paleolithic Mousterian.

Proto-Levallois and Mousterian Levallois cores bear many similarities, the most important of which is the shaping into two hierarchical surfaces. Additionally, both relied on the preparation of core convexities to produce predetermined blanks, on the creation of a ridge intersecting the two core surfaces acting as a striking platform, and on the removal of predetermined blanks in parallel with the plane of intersection. Moreover, in both cases, the cores were maintained by systematically removing éclats dèbordants flakes in order to control core convexities. This evidence suggests that the conceptualization of predetermined blank production—what has come to be known as the Levallois reduction concept—first emerged during late Acheulian times in the Levant.

Discoid cores are considered here as part of the parcel of prepared core technologies and are mostly characterized by a lack of hierarchy between core platforms (or having a switchable hierarchy), and by the removals of flakes at a typical obtuse angle. The discoid cores represented here differ somewhat from the classic Middle Paleolithic discoid cores, as a significant number of the scars cross the core midline, a feature not typical of later discoid cores. Prepared core CTEs, not commonly represented in Lower Paleolithic sites in the Levant, further add to our understanding of the complete reduction sequence. Prepared cores (general) reflect, in our view, a less distinctive technological procedure than the proto-Levallois and discoid. The technological analysis presented above suggests that this reduction sequence is predetermined, albeit less systematically, than the other two methods. A thin line separates this group of cores from the proto-Levallois ones. It is possible that cores defined by us as prepared cores (general) actually represent less successful attempts to execute the proto-Levallois method. Moreover, many of the CTEs defined by us as prepared core CTEs resemble débordant items, and were not classified as such only because they do not precisely fit the definition. With regard to the previously discussed difficulty of correctly identifying a variant that is on the periphery of a particular technology, the definition of this group as distinct from the proto-Levallois and discoid cores is intended as a possible methodological solution for such cases. We thus argue that proto-Levallois and prepared cores (general) should be viewed as a technological and conceptual continuum, reflecting early stages of the assimilation and adoption of predetermined blank production in the late Lower Paleolithic Levant.

The Revadim and Jaljulia cores were meticulously designed by both preparation removals and débordant removals, intended to facilitate the later production of desired predetermined items. The proto-Levallois reduction sequence is generally similar at both sites, excluding some differences that will be mentioned below. Notably, the proto-Levallois reduction sequence we defined here has been documented in other late Acheulian sites in the Levant and beyond (Gilead and Ronen, 1977; Goren, 1979; Ronen et al., 1980; Goren-Inbar, 1985; Chazan, 2000; DeBono and Goren- Inbar, 2001; Goren-Inbar, 2011; Shimelmitz et al., 2016; Zaidner and Weinstein-Evron, 2016; Goren-Inbar et al., 2018), and it thus can be seen as a hallmark of the late Acheulian lithic repertoire. In this paper we present the first comprehensive and detailed analysis of this intriguing phenomenon.

Although prepared cores and blanks represent a small percentage of the core and blank assemblages at both sites (Revadim: Agam et al., 2015; Solodenko 2010; Rosenberg-Yefet and Barkai 2019), we believe that their ubiquitous presence in every lithic assemblage at these two late Acheulian sites reflects the habitual use of prepared core technologies during this period in the Levant.

Proto-Levallois cores were more intensively shaped than discoid and prepared cores (general), as can be construed from the number of scars, scars representing débordant items, and cortex coverage. Discoid cores, however, were more systematically shaped around the circumference of the cores than the proto-Levallois cores, a difference that might be explained by their having been reduced along the entire core circumference, whereas reduction in proto-Levallois cores tended to be more restricted. The larger number of éclats débordants removals observed on the proto-Levallois cores might also explain this difference, as they might have spared the knapper the need to shape the entire circumference by means of bifacial flaking. Shaping the convexities of proto-Levallois cores by éclats débordants removals requires fewer bifacial blows at the lateral edges of the cores, as is common in Mousterian assemblages as well (Beyries and Boëda, 1983; Geneste, 1985:230; Meignen, 1995:280; Hovers, 2009:82; Shimelmitz and Kuhn, 2013; Hu et al., 2019). The use of débordant removals in order to design and maintain prepared cores in late Acheulian assemblages requires both expertise and an understanding of the blow direction, angle of striking, and precise use of force. The appearance of this technique as early as the late Acheulian demonstrates that knappers had the cognitive capabilities to master this distinctive and complex maintenance procedure long before its widespread adoption during Middle Paleolithic times.

The CTEs, included as part of our analysis of the entire reduction sequence and all its components, provide a glimpse into the initial shaping and maintenance stages of prepared cores: pre-planned core design and the strict adherence to concepts related to core volume and convexities. The presence of éclats débordants clearly demonstrates similarities in core design and maintenance with the fully-fledged Middle Paleolithic Levallois method, while éclats débordants a dos limité can be assigned, in the assemblages analyzed in this paper, to either the Levallois or the discoid methods.

Given the data presented above, the entire reduction sequence of prepared cores appears to be better represented in the Jaljulia sample than in the Revadim sample. This is reflected in the more frequent occurrence of both CTEs and blanks related to these core methods. Nevertheless, we have little doubt that the CTE items associated with prepared cores at Revadim also attest to the execution of core preparation and maintenance related to the Levallois concept. The high frequency of items characterized as éclats débordants in layer C3 East at Revadim, an assemblage analyzed by Malinsky-Buller et al. (2011), is noteworthy in this regard. The assemblage includes 209 éclats débordants and 56 outrepassés. Prepared cores (n = 61), or, in the authors’ terminology—cores with two surfaces perpendicular to each other with hierarchy—comprise 13% of all the cores in area C East. Intra-site variability in the application and intensity of prepared core technologies is of course to be expected, but nor can one disregard the presence of prepared core technologies at every single archaeological context at Revadim. We hope to gain a clearer understanding of intra-site variability once detailed analysis of all the areas at Jaljulia is complete.

The ridge angles of the éclats débordants items (Revadim: 62°; Jaljulia: 67°) differ significantly from the EPA angles (Revadim: 82°; Jaljulia: 83°). These angular differences reflect the asymmetry in the two platforms along the core circumference, a feature that generally characterizes proto-Levallois rather than discoid cores. Éclats débordants might thus have been detached mostly from the former type.

Inter-site variability can also be observed in the striking platforms of the éclats débordants items from the two sites. Whereas the Jaljulia items frequently exhibit faceted and dihedral striking platforms (éclats débordants: faceted, 45%; dihedral, 15%), these features are less frequent at Revadim (eclats débordants: faceted, 18%; dihedral, none). Even fewer of these faceted striking platforms (14.5%) were observed for the éclats débordants items in layer C3 East (Malinsky-Buller et al., 2011). In Mousterian sites, faceted or dihedral striking platforms are usually better represented (for example at Tabun; see Shimelmitz and Kuhn, 2013). The greater number of faceted striking platforms at Jaljulia is correlated with the more intensive treatment of convexities along the proto-Levallois core circumference (Tables 1, 4). The larger number of CTE items as well as the more intensive shaping of their striking platforms at Jaljulia might indicate greater assimilation of the Levallois concept there than at Revadim. The removal of these items was not symmetrical, as éclats débordants were repeatedly struck from one edge of the core: 81.6% were removed from the left lateral side at Revadim (Figure 10A) and 65% from the right lateral side at Jaljulia (Figure 10B). This asymmetrical treatment of core convexities is also common in other Middle Paleolithic Levallois assemblages in the Levant (Shimelmitz and Kuhn, 2013) and might reflect some technological continuity between late Acheulian proto-Levallois to the Middle Paleolithic Levallois.

As for blanks produced from prepared cores, the small number of these items at Revadim precludes a comparison between the two sites. Their infrequency is perhaps due simply to the difficulty in identifying these items. In this study, no distinction was made between proto Levallois flakes, discoid flakes and prepared cores (general) flakes. We are currently developing the methodology required to distinguish between these categories, to be published in future study. In a recent paper, González-Molina et al. (2020) used machine learning algorithms to address the challenges of distinguishing between discoid and Levallois flakes. Width at 50 and 75% of the flake length, inner angle, and maximum width were found to be the best parameters for accurate classification. The large number of faceted platforms in the Jaljulia assemblage is documented in many other Levallois assemblages, for example, at the Early Middle Paleolithic site of Misliya Cave, where convex faceted and “chapeau de gendarme” striking platforms are seen as diagnostic features of the Levallois method (Zaidner and Weinstein-Evron, 2020). These along with other factors will be considered in the future.

Changes in the Levantine habitat towards the end of the Lower Paleolithic might have played a role in the invention, adoption and assimilation of the Levallois method. The disappearance of elephants from the Levantine post-Acheulian landscape, marking the end of the Acheulian mode of adaptation, was probably a slow process, but one that must have had severe consequences for the Acheulian groups, who had lived in their presence for hundreds of thousands of years. The strong spiritual bond between early humans and elephants might be expressed in the production of replicas of the iconic handaxes from elephant bones, which seems to have accelerated towards the end of the Acheulian (Mussi, 2005; Costa, 2010; Zutovski and Barkai, 2016; Barkai, 2019; Barkai, 2021).

We suggest grouping together these separate lines of evidence into what seems to all be part of the same story: the assimilation and adoption of the Levallois method has its roots in the late Lower Paleolithic and might also have been driven by the disappearance of elephants from the Levantine habitat. This major habitat change may have created a “cultural storm” in which the adoption and assimilation of the Levallois method began in the late Acheulian, before the final extinction of the elephants, and became much more pronounced during the Middle Paleolithic. The AYCC, with its transition to the hunting and consumption of smaller prey, demonstrated a different kind of human adaptation and might well be part of the same story (e.g. Barkai et al., 2017).

The study of prepared core technologies at Revadim and Jaljulia provides an interesting observation point regarding the adoption and assimilation of technological innovations during Lower Paleolithic times. Technological innovations often involve a long and multi-participant process as the idea progresses and develops in what is known as “cumulative culture” (Tomasello, 1999; Hoppitt and Laland, 2013), also sometimes termed the ratchet effect (Tomasello, 1999; Tennie et al., 2009; Mesoudi, 2011). It has been argued that this accumulation of knowledge requires high-fidelity social learning mechanisms in order to allow the transmission of innovative ideas (Boyd and Richerson, 1988; Acerbi and Mesoudi, 2015; Tennie et al., 2016; Tennie et al., 2017). Some of us have previously argued that evidence for the existence of cumulative culture appears as early as the end of the Lower Paleolithic (Rosenberg-Yefet et al., 2021).

When considering the life cycle of a technological innovation, a few steps are generally expected: a slow inception, gradual assimilation of the technology, an increase in its frequency up to a certain peak, and then a decline, in an S curve representation, created by differences among individuals in their degree of “innovativeness” (although not all adoption curves are S-shaped) (Henrich, 2001). Five adopter categories are usually considered: innovators, early adopters, early majority, later majority, and laggards (Valente, 1996; Henrich, 2001). These differences between early and late adopters can create inter-individual variability regarding some technologies (Roux et al., 2018). This variability is expressed both in inter-group variability between individuals (Rogers, 2003; Correa, 2016; Assaf, 2021; Gandon et al., 2021) and, if we view the bigger picture in lower resolution, in inter-site variability along the chrono-geographical range. Many factors determine how innovative ideas and technologies are disseminated, including scales of mobility, population size, connectivity and networks, population stability, the presence of “weak ties”, defined as occasionally accessed connections (as opposed to “strong ties” such as family/kin), and environmental richness (Kuhn, 2012; Collar et al., 2015).

Given these high-resolution and low-resolution views of technological variability, we can see why a technology might be expected to differ in its initial versus peak phases. These differences will be expressed in the level of standardization and in the distinctiveness of the technology. Hence, we expect that some traits of a technology will be less standardized during the initial phases of adoption.

The Revadim and Jaljulia proto-Levallois assemblages reflect, in our opinion, an initial phase of technological adoption and assimilation. The partial correspondence of the proto-Levallois cores, CTEs, and prepared core end products to the Mousterian Levallois definition reveals the conceptual principles that will later define the technology at its peak phase.

The percentage of prepared cores among the total number of cores at Acheulian sites varies but remains a noteworthy phenomenon. At Berekhat Ram, for example, 64.8% of the cores are prepared (sample size is, however, restricted) (Goren-Inbar, 1985). At Holon, while only 3.2% of the cores are prepared, this percentage increases to 19.3% for cores with hierarchy, which might correspond to our definition of prepared cores) (Malinsky-Buller, 2014). At the different layers of Gesher Benot Yaʻaqov, 5.2–17.3% of the cores are prepared (Goren- Inbar et al., 2018). These are but a few of many examples of assemblages that might reflect early stages of adoption of the technology, as suggested for the Revadim and Jaljulia cases, and therefore can be seen as part of the chrono-geographical range of inter-site variability, mentioned earlier.

The early appearance of prepared cores at both sites signals, in our view, the inception of concepts related to the Levallois method, termed here proto-Levallois, in the late Acheulian Levant. The technological and conceptual characterization of the initial stage presented here reflect a broader behavioral model of adoption and assimilation of technological innovations during Lower Paleolithic times.