Introduction

The process of Neolithization in Southwest Asia is viewed as a major transition in human history, encompassing sedentism, the emergence of agriculture, animal husbandry, demographic growth and new forms of social organisation. In search for the origins and drivers of such fundamental processes manifested by the early Neolithic record, starting ca. 11,700 cal. BP, scholars turned their attention to the preceding period known in Southwest Asia as the Epipalaeolithic, divided into three phases: Early, Middle and Late. Based on techno-typological characteristics of microlithic assemblages this division has been associated in the Mediterranean Southern Levant with three major cultural entities: the Kebaran (ca. 25,000/24,000–18,500/18,000 cal. BP), Geometric Kebaran (ca. 18,500/18,000–15,000 cal. BP) and Natufian (ca. 15,000–11,700 cal. BP) (Bar-Yosef, 1970; Goring-Morris & Belfer-Cohen, 2011; Goring‐Morris & Belfer‐Cohen 2017; Richter & Maher, 2013).

Since its definition by Garrod (1932), the Natufian culture has played a central role in the study of the origins of the Neolithic (Bar-Yosef, 1998; Goring-Morris & Belfer-Cohen, 2011; Grosman & Munro, 2017). It is commonly accepted that even though Natufian sites do not show direct evidence for domesticated crops and livestock, they do present solid evidence for reduced mobility, and possibly sedentism (Bar-Yosef & Belfer-Cohen, 1989; Grosman et al., 2016; Stutz et al., 2009; Valla, 2018; Yeshurun et al., 2014a, 2014b). This is largely based on the faunal record showing evidence for commensal house mice remains found within Natufian habitation sites (e.g., Tchernov, 1984; Weissbrod et al., 2017), and the transition to a broader spectrum of resources argued to be associated with the exhaustion of local large and medium game resources due to continuous and intensified occupation (Munro, 2004; Stutz et al., 2009; Yeshurun et al., 2014b). In addition, many Natufian sites are characterised by durable structures with stone foundations (e.g., Edwards, 2013; Grosman et al., 2016; Valla et al., 2007; Yeshurun et al., 2014a), burial grounds (e.g., Belfer-Cohen, 1988; Bocquentin, 2003; Byrd & Monahan, 1995; Friesem et al., 2019; Grosman et al., 2016; Nadel et al., 2013; Noy et al., 1973; Perrot & Ladiray, 1988; Weinstein‐Evron, 1998, 2009), large installations interpreted as storage facilities (e.g., Grosman et al., 2020; Perrot, 1966), large and heavy grinding and pounding stone tools (e.g., Belfer-Cohen & Hovers, 2005; Dubreuil, 2004; Rosenberg & Nadel, 2014) and high density of finds in the occupation deposits (Arranz-Otaegui et al., 2018; Grosman et al., 2016; Hardy-Smith & Edwards, 2004; Valla et al., 1991; Valla et al., 2007; Weinstein‐Evron, 1998, 2009; Yeshurun et al., 2014a, 2014b).

The earliest Mediterranean Epipalaeolithic entity in the southern Levant, the Kebaran, does not seem to express a major departure from Upper Palaeolithic traits as evidenced by the intensity of occupation at the temporary camp sites, hunting strategies, and exploitation of plants (Marom & Bar-Oz, 2008; Stiner, 2005), although the Ohalo II fisher-hunter-gatherer site provides a different example (Nadel, 2017; Nadel et al., 2004; Snir et al., 2015; Steiner et al., 2022). The Middle Epipalaeolithic entity, known as the Geometric Kebaran (hereafter GK), however, has been argued to exhibit a continuation of Early Epipalaeolithic subsistence strategies but with increasing intensity of occupation (Kaufman, 1992; Liu et al., 2020; Maher, 2017). Kaufman (1992), based on his excavations at the GK site of Neve David that revealed graves, groundstone implements, a stone-built feature and deep deposits over an area exceeding 1,000 m2, suggested that the site’s salient features foreshadow the Natufian and reflect similar mundane behaviours at the camp, though at a lesser scale. He suggested that relatively large-scale aggregation events were taking place at the site, and thus the site should be viewed as a precursor to the Natufian permanent aggregation phenomenon. Based on these studies, and relying mainly on the density of lithic assemblages, Maher (2019) further suggested viewing the major Kebaran and GK sites as large aggregation locations bringing together populations from different groups to engage in some sort of social interaction. However, as put forward by Conard et al. (2012), cases showing an increase in lithic density over time at the same site could be the outcome of two very different scenarios: (1) an increase in the number of people occupying the site, while the duration of occupation might remain similar (i.e., relatively short); (2) a longer duration of occupation, while the population size remains similar (i.e., relatively small). Thus, it remains an open question which of the two scenarios is reflected in the large GK sites, or if there was a combination of the two. The relation between sedentism and an increase in the population size occupying the site has been long argued (Belfer-Cohen & Bar-Yosef, 2002; Bocquet-Appel & Bar-Yosef, 2008; Dow & Reed, 2015; Rosenberg, 1998), suggesting that even in the cases where reduced mobility started with small sized groups, it would eventually result in demographic increase. Furthermore, these processes – broadly termed as the transition to sedentism – allegedly starting in the GK with increased occupation intensity – had a profound impact not only on demography but also on cultural organization and innovation.

Neve David, one of the key GK sites in the Southern Levant, has played a major role in the conceptualization of increased occupation intensity in the GK. The results of Kaufman’s excavations at the site in the 1980s have been repeatedly employed to argue for the development of large-scale aggregations, or merely more intensive occupations, prior to the Natufian (Delage, 2018; Goring‐Morris & Belfer‐Cohen, 2010; Kaufman, 1989, 1992; Maher et al., 2012). In order to better characterize and understand the nature of increased GK occupation intensity, renewed excavations at Neve David were carried out between 2014–2019 aiming to reassess the integrity, formation and modification of the thick and dense GK deposits. Here we present the study of archaeological formation processes at the site and discuss their implications for reconstructing and understanding occupation intensity. To do so, we apply a multi-scalar spatial approach that considers taphonomic issues and diachronic changes in space use by analysing lithic, faunal and micro-botanical assemblages, alongside a micromorphological study of the site's sediments. Our working hypothesis is that the thick layers at the site are the result of human occupation and not the outcome of post-depositional processes. If we are correct, then the GK record at Neve David can be used in cultural interpretations, including testing for socio-economic changes indicative of increasing occupation intensity. However, if the deposits are mostly the outcome of post-depositional processes, then Neve David should not be an anchor for models postulating increased occupation intensity in the Middle Epipalaeolithic.

The site of Neve David

Neve David is a large open-air site on the western foothill of Mount Carmel, Israel, within the modern city of Haifa. The site occupies an ecotone between the mountain range and the Mediterranean coastal plain, and is on the northern bank of Nahal Siah on its outlet to the plain (Fig. 1). Currently the coastal plain west of the site is just one km wide, but it was much wider in the late Pleistocene (Sivan & Greenbaum, 2017). The site was discovered by H. Laville during road construction in 1983, and then excavated on behalf of the University of Haifa by D. Kaufman for four seasons between 1984 and 1990, during which large numbers of lithics and animal bones were unearthed. The main archaeological layer (characterised by a red-brown clay known as terra rossa, common on Mount Carmel) was assigned to the GK culture of the Middle Epipalaeolithic period based on the typo-technological characteristics of the flint assemblage, where trapeze/rectangle microliths are abundant. Two primary burials associated with ground-stone implements were excavated (Bocquentin et al., 2011; Kaufman, 1988; Kaufman & Ronen, 1987; Yeshurun et al., 2015). Two radiocarbon dates derived from burned animal bones associated with one burial yielded an age range of ca. 16,700–14,200 cal BP (Bocquentin et al., 2011; Kaufman, 1987, 1988, 1989). The large faunal assemblage indicated hunting, butchery and consumption of mountain gazelle (Gazella gazella), Mesopotamian fallow deer (Dama mesopotamica), and occasionally smaller game (Bar-Oz et al., 1999).

Fig. 1
figure 1

Map of Mount Carmel region, including Neve David and other sites mentioned in the text

Full size image

Beginning in 2014, D. Nadel and R. Yeshurun on behalf of the Zinman Institute of Archaeology, University of Haifa, resumed excavations at the site (Fig. 2). Based on five seasons of excavation in the main area, and the profiles of five trenches, the thickness of the GK deposits is ca. 1 m across most of the site; the size of the site, while only partially preserved, was at least 1,600 m2 (Yeshurun et al., 2015). The flint assemblage is rich, mostly utilizing the outcrops on the slopes above the site and focusing on the production of bladelets and microlithic tools (Liu et al., 2020). Faunal remains are common and in the deeper parts large skeletal elements are well preserved. Across the site, the GK deposits are cut in numerous places by pits and other features dating to the Pottery Neolithic (Shochat et al., 2019). However, the different colour and texture of the post-GK sediments, and their different cultural contents, make it possible to separate “clean” GK deposits from later or mixed ones (Figs. 2, 3a); only the former are discussed here.

Fig. 2
figure 2

A map of the renewed excavations at Neve David. Red-brown colour denotes the GK sediments: a) plan of the excavation area, bold numbers denoting loci; b) a typical section showing the GK deposits, with no clear floors or trampled surfaces visible, and the remains of an overlying stone-built Neolithic feature (Locus 1)

Full size image
Fig. 3
figure 3

The renewed excavations at Neve David, field photos: a) Level XIV at the end of excavation, looking north-west. The GK layer appears as a brown sediment, cut in places by the Neolithic pits and a stone structure; b) a portion of Level II, showing Locus 5, with a distinct concentration of bones and lithics; c) articulated gazelle incisor teeth in Level II (marked by an arrow); d) conjoined gazelle rib fragments, Level II; e) a complete gazelle scapula, as exposed in Level XII

Full size image

No GK stone-built features or graves were found in the renewed excavations, however we found numerous small pits, often with large densities of lithics and animal bones (Fig. 3). One such feature is Locus 5 (Fig. 3b), which included some butchered gazelle and deer bones as well as 90 microliths that were mostly used as insets of composite projectiles (Groman-Yaroslavski et al., 2020). The 1-m-thick deposits in Trench VI was the subject of a study that incorporated lithic densities and technological indices as a proxy for site-occupation intensity along the stratigraphic sequence, concluding that there was a rise in occupation intensity over time (Liu et al., 2020).

Materials and Methods

Field Excavations

The excavation was performed using a general 1 × 1 m grid, in basic units of 0.5 × 0.5 × 0.05 m, with basket numbers assigned to each excavation unit. In some cases, a 0.5 × 0.5 × 0.05 m sub-square was excavated in two or more baskets, when lateral alterations in the sediment or density of finds were observed. Features such as stone arrangements, special sediment patches or concentrations of finds were designated as loci and were excavated in the aforementioned grid system. The sediment volume of each basket was recorded in the field. All baskets underwent wet sieving with 2 mm and 1 mm meshes. We aimed at excavating wide areal exposures, within the limitations imposed by the Neolithic and recent intrusions. Each such exposure, typically 2–5 cm thick and following the natural (gently sloping) stratigraphy, consisted of many stratigraphically-related baskets and was designated as a “level” and denoted with a Roman numeral (Levels I–XV; the numbers increasing following the order of excavation). For each exposed level, we systematically sampled the sediments for phytolith and bulk sediment analyses, and recorded the orientation and inclination of elongated flint and bone items. The levels vary in their areal extent due to the post-GK intrusions in the upper levels, and we primarily focused our excavation on the richer and seemingly better preserved squares in the northern part of the exposed area, especially in the lower levels.

Overall, the 15 levels composing the long stratigraphic sequence now available at the site seemed quite similar when encountered in the field. In order to examine the temporal and spatial changes in patterns of material deposition throughout the long GK sedimentary sequence we decided to focus this study on three representative sets of levels: (1) II-III from the top part of the sequence; (2) VI-VII from the middle of the sequence; and (3) XII-XIV from the bottom of the sequence. For each set of levels we conducted spatial analysis of three major archaeological finds: lithics, bones and micro-remains, to facilitate the study of diachronic changes during the GK occupation at Neve David.

Lithic Analysis

Each basket of the flint assemblages from the three level sets was weighed, and the total per level calculated and divided by excavation volume to obtain a lithic density (by weight) per square. Furthermore, this study focuses on one knapping product, cores. The study of the entire lithic assemblage is beyond the scope of this article, and hence cores were selected to represent certain aspects of the lithic assemblage. They are numerous, easily identified, and an adequate type for analysis when weight and signs of exposure to fire are concerned (it is easy to identify burning signs on cores, and sometimes more challenging on smaller elements such as bladelets). The identification of lithics exposured to fire is relevant as these attest to the presence of hearths. A higher number of hearths, in turn, reflect higher number of visits to a site or higher occupation intensity (Friesem et al., 2017, 2024; Mallol et al., 2007). Thus, burnt cores are considered a reliable proxy for the presence of hearths. The changes in bladelet core frequencies are tied here to the issue of ad-hoc vs. curated technology. Bladelet cores are considered a proxy of curated technology, as they have a pre-planned morphology with specific methods of preparation and maintenance. Following ethnographic examples, we expect that the more sedentary a group is, the less investment in core preparation would be evident (Parry & Kelly, 1987). Furthermore, as bladelet cores form the majority in each level, and as they reflect a main production target (i.e., bladelets and hence the microlith assemblage) throughout the sequence, their percentages and volumetric densities are used in the following analyses as proxies for hearth presence at the sampled levels (see also Liu et al., 2020, p. 3).

Furthermore, in a previous diachronic lithic study of the assemblage from the deepest trench (Trench VI; Liu et al., 2020), the cores in general and the bladelet cores in particular showed a diachronic trend through the stratigraphic sequence. Hence, we studied all the cores from Levels II-III (n = 182), Levels VI-VII (n = 150), and Levels XII-XIV (n = 502) and the frequency of bladelet cores in each, as well as the frequencies of burnt cores that likely represent the intensity of camp use (the presence of hearths and their impact on material remains, and see below). The fabric analysis focused on 294 elongated flint items from these levels (Appendix 1).

Faunal Analysis

Here we present a density analysis of the faunal record. All bone specimens were collected from the 2 mm sieve and saved for further analysis. Each bone bag, representing a single basket, was weighed and the result was divided by excavation volume. The fabric analysis employed 106 elongated bone items from the studied levels (see Sect. "Spatial and fabric analyses" and Appendix 1).

Micro-Geoarchaeology

Bulk sediment samples were collected during the excavation seasons in September 2016 (n = 40), 2017 (n = 22) and 2019 (n = 30) and one sediment sample was collected in 2015 outside the site to serve as a control sample for regional soil characteristics. Bulk sediments were sampled from the GK layers exposed during the excavation in order to spatially cover the new excavation area. In addition, selective sampling was applied to specific loci of interest according to field observations.

All bulk sediment samples were analysed using Fourier-Transform Infrared (FTIR) spectroscopy in order to identify the major mineral and organic components. Some samples were divided to sub-samples in order to separate different types of sediment, bone, flint, shell and other materials within each sample. In total 138 samples were analysed. Infrared spectra were collected using the KBr method between 4000 and 250cm−1, at 4cm−1 resolution using a Thermo iS5 spectrometer following Weiner (2010) and interpreted using the Kimmel Center for Archaeological Science Infrared Standards Library, Weizmann Institute of Science (https://centers.weizmann.ac.il/kimmel-arch/infrared-spectra-library).

The concentration of silicate phytoliths, an indicator of the presence of micro-botanical residues, and the concentrations of wood ash particles, were quantified from a representative set of the bulk sediment samples (n = 51) and the regional soil control. Extraction and quantification of phytoliths followed the method described by Katz et al. (2010). Phytolith concentrations were then calculated in millions per 1 gr sediment [M phyt/gr sed] with 30% standard deviation. Wood ash particles were calculated by quantifying calcitic wood ash pseudomorphs following the method described by Gur-Arieh et al. (2013). Wood ash pseudomorph concentrations were then calculated in thousands per 1 gr burnt sediment [K ap/gr sed] with 30% standard deviation.

Undisturbed monolith sediment blocks (n = 5) were collected for the preparation of thin sections for micromorphological analysis. Blocks were collected from profiles exposed during excavation of the GK layers. A total of five thin sections were produced at the McBurney Laboratory for Geoarchaeology at the University of Cambridge following the method described by Murphy (1986). Thin sections were first studied at a scale of 1:1, scanned using a flatbed scanner, and then analysed with a petrographic microscope at magnifications ranging from 4 × to 400 × with plane-polarized light [PPL] and cross-polarized light [XPL]. Micromorphological descriptions employ the terminology of Stoops (2003). Quantification of micro-bone fragments (< 2mm) was calculated using a 1mm grid laid on the thin section. Micro-bone fragments were counted in 30 fields in order to calculate the density of bone per cm2.

Spatial and Fabric Analyses

We conducted fabric analysis (e.g., Giusti et al., 2019) on all elongated flint and bone items in the studied archaeological level sets, which were recorded following the exposure of each level. We plotted the orientation (direction) of the items, hypothesizing that on undisturbed living surfaces the orientations (north–south, northeast-southwest, southeast-northwest and east–west) would be randomly divided, while deposition or redeposition by natural agents (i.e. fluvial) would produce (a) dominant orientation/s. Additionally, we examined the inclination (dip) of the artefacts relative to the horizontal plane, assuming that an undisturbed surface would display mostly planar (0°) inclinations. The recording employed 0, 30, 45, 60 and 90 degree categories. We tested for uniformity of the orientations and inclinations with Rayleigh's test for uniform distribution in PAST version 4.03 (Hammer et al., 2001). As these two aspects were studied together for both materials, the results are presented together.

Georectified aerial images were produced from three-dimensional Structure from Motion (SfM) models of the excavation made in 2016, 2017 and 2019 using Metashape Pro (e.g., Willis et al., 2016). These plans were georeferenced according to the local excavation grid, and archaeological features and excavation areas were traced manually from the georeferenced images in Adobe Illustrator CC 2022 (https://www.adobe.com/products/illustrator.html) using the Avenza MAPublisher 11.1 GIS plugin (https://www.avenza.com/mapublisher) to produce shapefiles. These shapefiles were used to illustrate the distribution maps (described below).

Spatial maps were generated using QGIS 3.26 (https://www.qgis.org). Major grids (1 × 1 m) and sub-grids (0.5 × 0.5 m) were drawn over the excavation area using the Create Grid function and were manually linked to the excavation data. Volumetric choropleth maps are presented by weight according to artefact density (g/m3) for lithics and bone. Micro-remain data (phytoliths and ash pseudomorphs) from bulk sediment sample data are presented by subsquare according to the concentrations per gram of sediment. Note the spatial maps for each artefact or micro-remain share the same legend and data breaks to facilitate comparison between levels. Local grid coordinates (x, y, z) were taken for bulk sediment samples using a total station in 2016 and 2017, and these points are indicated on the micro-remain maps.

Results

Lithic Densities and Spatial Analysis

The total density of flint artefacts (by weight) significantly decreases with depth, though with some fluctuations (Fig. 4a: Spearman’s r = -0.81, p < 0.01). The spatial distribution of lithic items is overall quite uniform and does not attest to regular deposition (or redeposition) of flint items in preferred locations through time. Levels XII-XIV display a somewhat higher density of material (total lithics as well as cores) in the northwest part of the excavation area, but this pattern is not clearly repeated in other levels.

Fig. 4
figure 4

Artefact densities (by weight) through the stratigraphy of Neve David: a) total lithics; b) total bones. The three studied level groups are coloured in dark blue.

Full size image

A total of 922 cores were studied here (Table 1). The density of cores somewhat increases from Levels XII-XV to Levels VI-VII, and increases dramatically in the upper levels. The frequencies of bladelet cores out of all cores (signifying curated technology) increases from 66.3 % Levels XII-XV to 77.6% in Levels VI-VII, and then decreases towards the top (69.2%). This is a statistically significant difference (χ2 = 6.70, p = 0.04, although with a high effect size, as Cramer’s V = 0.08), stemming from the higher proportion of bladelet cores in the middle levels. Burnt core proportions are similar throughout the sequence (18–19%; χ2 = 0.22, p = 0.89).

Table 1 Core types, burning signs and densities (m3) in the three studied assemblages
Full size table

In some parameters there is a clear trajectory through time, where the densities of cores, the densities of bladelet cores and the frequencies of burnt bladelet cores increase through time, while in terms of frequencies of bladelet cores and frequencies of burnt cores in the middle levels there is a deviation from the former pattern (Table 1).

The spatial patterning of the lithic assemblage (total weight per volume) shows some vertical trends (Fig. 5a). The northern half of the upper levels seems quite homogeneous, while the southern half has several squares with high densities. The middle levels are similar to the upper ones, albeit with only one high-density square. The lowest levels are again quite homogenous, with somewhat higher densities in the northwest.

Fig. 5
figure 5

The spatial distribution of flint items in each level, shown as lithic weight divided by excavation volume in each grid subsquare: a) total lithics; b) cores; c) burnt cores

Full size image

The distribution of the cores (Fig. 5b) is somewhat similar to the distribution of the assemblage as a whole. In the upper levels there are higher densities in the south, the middle levels are irregular and the bottom levels are denser, with a cluster in the northwest. The same is true for the burnt cores. In sum, the bottom levels are quite homogeneous with higher densities in the northwest; the middle levels are also homogenous regarding the total assemblage, while the cores show several dense clusters; and the top levels are irregular with higher densities in the south.

Faunal Spatial Analysis

The weight densities of faunal remains significantly decrease with depth, though the trend is not as strong as with the lithics (Fig. 4b; r = -0.61, p = 0.02). The sample from Level XIV, the lowest widely exposed level, stands out as presenting a high density of faunal remains. The spatial distribution of bones does not clearly indicate temporally consistent spatial concentrations of material. In the lowest levels (XII-XIV), more and/or heavier bones were located in the northwestern part of the excavation area, while the other parts of the area display a poorer and uniform density of bones. In the overlying levels, several sporadic concentrations were found in the center, northeast and south (Fig. 6). Nevertheless, the faunal remains in each sample display evidence of minimal movement, such as the presence of conjoined rib fragments and articulated incisor teeth of a gazelle in Level II, an articulated piece of a tortoise plastron in Level VI, a cluster of gazelle bones (mandible with teeth, articulated unfused distal metapodials and a phalanx) in Level VII, and horizontally laid and complete gazelle and deer scapulae in Levels XII-XIV (Fig. 3c–e).

Fig. 6
figure 6

The spatial distribution of faunal remains in each level, shown as bone weight divided by excavation volume in each grid subsquare

Full size image

Fabric Analysis

The elongated flint and bone items demonstrate a generally uniform distribution in four out of six cases (i.e. flint or bone in each of the three tested levels; Fig. 7a), meaning that major fluvial or other winnowing agents did not redeposit the archaeological material. The flint items in Levels VI-VII and bones in Levels XII-XIV possibly diverge from this pattern, as their Rayleigh's test rejected the hypothesis of uniform distribution (p-value < 0.05); however, this result was not obtained by the other material type (bones and lithics, respectively) in the same levels. The inclination angles of the same items relative to the horizontal plane are significantly concentrated around 0° (planar) in all samples (Fig. 7b; see statistical test results there). More lithic items were found to diverge from the planar dip than bones, perhaps because the mechanical properties of bone items make them more amenable to trampling and aligning horizontally as a result of human traffic. In any case, the strike and dip patterns we recorded do not show great differences between the levels, and generally rule out any significant movement or disruption of the elongated materials.

Fig. 7
figure 7

(a) The orientations of elongated flint (upper row) and bone (lower row) items, presented as azimuths (north = 0) in rose diagrams for each level. Rayleigh's R tests and p-values are given for each diagram. (b) Inclination angles (relative to the horizontal plain) of the elongated flint (upper row) and bone (lower row) items (0° = planar) for each level. Rayleigh's R tests and p-values are given for each diagram

Full size image

Micro-Geoarchaeology

The majority of the samples present similar mineralogical composition, with clay and quartz as the main mineral components alongside calcite (Fig. 8a; Appendix 2). Examining changes in quartz density, size, sorting and shape do not show any significant difference between the layers, supporting a relative similar sedimentation rate across the archaeological sedimentary sequence. Most sediment samples from the GK layers show no indication for mineral alteration due to exposure to high temperatures (> 500°C) (Berna et al., 2007). However, four samples concentrated at the southern part of the excavation area, all from Level III (ND 16–30, 32, 33 and 37) contain dark color aggregates in the sediment. When measured separately from the general sediment, the aggregates indicate alteration of the clay component due to exposure to elevated temperatures (Fig. 8b).

Fig. 8
figure 8

Representative infrared (FTIR) spectra. (a) General brown GK sediment showing clay (1032, 528, 914, 3622 and 3696 cm−1), calcite (1438, 873 and 713 cm−1) and quartz (shoulder at around 1080 and 797 cm−1); (b) The spectrum of a dark clay aggregate from the southern part of the excavation area (ND 16-33b) showing altered clay due to exposure to high temperature (note the absence of absorbance bands at 525, 914, 3694 and 3622 cm−1), calcite and some quartz

Full size image

In terms of phytolith concentrations, most of our samples did not differ from the off-site control, but there are several distinct spots with higher concentrations. Based on the phytolith concentration in the regional control sample (ND 15–110: 0.23 M phyt/gr sed) and taking into consideration a 30% standard deviation (Katz et al., 2010), we divided the results of the archaeological samples into three groups: (1) < 0.5 M; (2) 0.5–1 M; (3) > 1 M phyt/gr sed; in order to distinguish between a natural or low signal of human activity and clear signals for human activity (Appendix 2). Out of 51 samples, only three samples (ND 16–11, 15 and 43) in Levels II-III, two samples (ND 17–28 and 30) in Levels VI-VII and one sample (ND 19–16) in Level XIV show moderate amounts of phytoliths, ranging between 0.5 and 0.9 M phyt/gr sed, indicating human exploitation of plants at these localities. High concentrations of phytoliths (> 2 M phyt/gr sed) are found in two samples (ND 19–6 and 20) in Levels XIII and XIV respectively. Both samples are from distinct features that were noticed during fieldwork; the former from Locus 47, an oval patch of compact yellowish sediment (ca. 50–100 cm across) defined by small stones, and the latter from Locus 49, a pit of yellowish sediment surrounded by small gravel (ca. 30 cm across) (Fig. 9a).

Fig. 9
figure 9

Spatial distribution of micro-remains in each level: a) phytoliths; b) wood ash pseudomorphs. The bold line represents the excavation boundaries for each level. Numbers in red indicate the locations of loci (features) mentioned in the text

Full size image

Similarly, distinct wood ash concentrations are uncommon and not randomly distributed. Based on previous studies (Friesem et al., 2014; Gur-Arieh et al., 2013) a conservative threshold for the concentration of wood ash indicating in situ human pyrogenic activity, was set at 500 K ap/gr sed. Since the regional control sample (ND 15–110) yielded 7 K ap/gr sed, a middle range category signalling a moderate signature of human activity (either as dispersed ash or mixed deposits) was set between 100–500 K ap/gr sed (Appendix 2). Out of 21 samples from Levels II-III, 11 samples show wood ash concentrations above 100 K ap/gr sed and two samples from Locus 5 (ND 16–10 and 12, Level II) show high concentrations with 949 and 836 K ap/gr sed, respectively (Fig. 9b). Given the high density of flint and bones in Locus 5, including 90 microliths and butchered bones, alongside the elevated concentrations of fire residues, it is possible that Locus 5 represents an in-situ feature, and less likely – a waste area. In Levels VI-VII, out of 16 samples, five samples yield wood ash concentrations between 106–212 K ap/gr sed, while the rest present values below 100 K ap/gr sed. It is difficult to determine whether the concentration of moderate levels of ash in squares U-T21 represents a palimpsest of a combustion feature. In Levels XII-XV, out of 15 samples, only three samples (ND 19–2, 8 and 10) yield wood ash concentrations between 113–212 K ap/gr sed (the latter sample is from a small pit feature, Locus 48), while the rest present values below 100 K ap/gr sed.

The results of micromorphological analysis of thin sections (Table 2) show that the brown sediment associated with the GK occupation is characterised by calcitic-clay groundmass with crystallitic b-fabric. The coarse/fine (c/f) related distribution is porphyric showing a poorly sorted and complex microstructure with mainly channels, chambers, vesicles and vughs voids (Fig. 10a and b). The sediment is only slightly disturbed due to biological activity (e.g., roots, rodents, insects, etc.), alongside indications for some water movement within the sediment. The southern part of the new excavation area (grid squares T–X/16–17) shows a more open and porous structure due to increased bioturbation in comparison with the northern part (T–X/18–23) which seems more intact and less disturbed (Fig. 10c).

Table 2 Micromorphological description and interpretation of thin sections
Full size table
Fig. 10
figure 10

Soil micromorphology of the general brown GK sediment (in all the microphotographs, the left photograph was taken in plane polarised light (PPL) and the right photograph in crossed polarised light (XPL)): (a) Scan of a thin section (NDB 17–3) from the northern part of the excavation area showing a crumbly poorly sorted structure with large fragments of bone and flint; (b) Microphotograph of the sediment from the northern part of the excavation area (NDB 16–3) showing the complex microstructure with many channels. Note the high abundance of micro-bones in PPL and the formation of micritic calcite around voids due to root activity in XPL; (c) Microphotograph of the sediment from the southern part of the excavation area (NDB 16–5) showing an open and porous structure indicating disturbance due to intensive bioturbation; (d-e) microphotograph 17–3/18–2 showing fire residues

Full size image

Throughout the different levels spatial patterns are quite similar, with residues of human activity in the form of tiny bone (with some of them showing signs of burning) and flint fragments; micro-charcoal and ash particles are present in varied amounts but overall with higher abundance in the northern part of the excavation area (Fig. 10e and f). The quantification of micro-bone fragments (< 2 mm) (Table 2) indicates higher abundance in the north-west part of the site (NDB 16–3 and 7; NDB 17–3 and 4; NDB 18–2 with values between 11–28 B/cm2) as opposed to the southeast area (NDB 16–5 with 6 B/cm2).

Discussion

Taphonomy and State of Preservation

Mineralogical analysis of the sediments at Neve David does not show any evidence for chemical alteration in the form of authigenic minerals, as found in the nearby Mount Carmel Palaeolithic cave sites where they indicate chemical diagenesis and alteration of the original archaeological record (Friesem et al., 2021, 2022; Karkanas et al., 2000; Shahack-Gross et al., 2004; Weiner et al., 2002). Micromorphological observations suggest only minor to moderate bioturbation within the sediments, mainly concentrated in the southern part of the new excavation area. The high degree of uniformity seen in the orientation of elongated items and their mostly planar dip, the concentration of specific finds (ash and phytoliths, burnt cores, and faunal remains displaying minimal movement), as well as the preservation of numerous small pits that are still distinguished macroscopically, indicate that the distribution patterns of occupation deposits at Neve David should be regarded as a result of human activity and much less so due to post-depositional natural processes (see discussion on anthropogenic vs. natural formation processes in Shahack-Gross, 2017). Given this site-formation evidence, the first of its kind from the site, we can now turn to the aspects of space use and occupation intensity at Neve David.

Use of Space

The deep sedimentary sequence associated with GK material culture exposed at Neve David, coupled with our study, suggest that the site is composed of multiple events of occupation. Unfortunately, attempts to date the sequence have not been successful so far. The archaeological record at Neve David includes almost no charcoal and no collagen is preserved in the bones. In addition, the GK layers do not overlie Kebaran (early EP) deposits and are not sealed by Natufian (late EP) layers. Importantly, for some reason this differs from nearby cave sites such as Raqefet (Lengyel, 2007), and the multi-layered Epipalaeolithic Ein Qashish South site to the east of Mount Carmel (Yaroshevich et al., 2016), where Kebaran, Geometric Kebaran and Natufian remains are found at the same site. In addition, the techno-typological characteristics of the flint assemblage do not exhibit significant change throughout the layers and (again) do not contain any Kebaran or Natufian elements (Liu et al., 2020). Thus, this sequence may represent any span of time between a few hundred and 3,000 years.

In a previous study of Neve David, it was shown that the density of lithics along the sequence exposed in Trench VI increases with time, which has been interpreted to evince an increased intensity of site occupation (Liu et al., 2020). No clear layers of occupation have been identified at the site, although the clusters of remains (some in pits), as well as the lithic, faunal and microscopical remains strongly suggest human occupation events and minimal post-depositional disturbance. Thus, the question remains whether the multiple episodes of occupation represent (1) increased mobile group size occupying the site for short periods of time in each event; (2) a similar (relatively small) group size but with reduced mobility and longer duration of occupation at each event; (3) a more complex scenario that represents a combination of the two. While such a question might not be unequivocally answered by our study, we hypothesize that, at least to begin with, the occupation time was prolonged, relative to earlier Epipaleolithic groups in this environment, and only later was there an increase in group size. We base this on our integrated micro- and macro-analyses, ethnoarchaeological data and comparison to other Palaeolithic sites in the Southern Levant.

In spite of the good preservation of camp structure and the evidence that the macro- and micro-artefacts are largely in situ (see above), the spatial distribution of the lithics, bones and micro-botanical residues do not show a clear pattern of division of space into specific activity areas. While microscopic fire residues (e.g., wood ash pseudomorphs, charcoal, burnt bones and flint) are found in high concentrations sporadically across the site, attesting to human use of fire, nowhere could we identify an in situ GK combustion feature or a clear ‘floor’. While there is clear evidence for the use of plants (in the form of high phytolith concentrations), on-site flint knapping and processing of faunal remains, we could not identify specific areas where such activities took place exclusively. However, it seems that one of the only features preserved is a refuse area associated with Locus 5.

In an ethnoarchaeological study focusing on the relation between mobility patterns and the archaeological signature of space use among highly mobile societies, Friesem et al. (2024) compared patterns of space use and archaeological formation processes among three contemporary societies. The three studied societies live today in tropical forests and show similar social characteristics (e.g., being small-scale groups, egalitarian, practicing sharing and traditionally classified by anthropologists as hunter-gatherers). However, due to different circumstances, each group presents a distinct pattern of mobility, occupying their habitation sites for different lengths of time: days to weeks, seasonal, and permanent. They show that only the group who occupied the site for a few days and up to few weeks left an archaeological signature of their activity areas. Groups who occupied their habitation site for over a month started to maintain their site by removing in situ activity residues from their primary location and disposing of them in waste areas. This practice significantly reduced the chances to archaeologically understand the use of space based on spatial distribution of primary activity remains after the abandonment of the site. The longer the site is occupied, the division of space (this time based on secondary refuse deposition) between domestic areas and waste areas becomes clearer, and chances for the preservation of primary activity remains in the domestic area are less likely (Friesem & Lavi, 2017, 2019; Friesem et al., 2017, 2024).

We therefore conclude that the Neve David occupations were not highly ephemeral in nature; people were performing different tasks over time in the same locales but repeatedly (and most likely unconsciously) trampling and disturbing the activity remains deposited on the surface of the camp. These findings are in contrast to earlier Palaeolithic open-air sites (e.g., Nesher Ramla or even Ohalo II (see Friesem et al., 2014; Nadel, 2017 respectively), and they may indicate some departure from ephemeral occupation during the GK of the Mediterranean Levant. Furthermore, the findings from Neve David are common in many later Natufian sites. At Baaz rockshelter, Stahlschmidt et al. (2017), report an in-situ hearth and a trampled floor they observed microscopically. But this seems to be the exception rather than the rule. When looking at micromorphological studies of Natufian open-air sites with a long occupation sequence such as el-Wad Terrace (Tsatskin et al., 2017), Wādī Ḥammeh 27 (Prossor, 2022), Eynan-Mallaha and Nahal Ein-Gev II (Friesem, personal observations and internal reports), all seem to represent a more sedentary activity; trampled floors were generally not identified and the very few fine horizontal laminae suspected to be floors are highly disturbed (e.g., Prossor, 2022). The lack of clear division of space throughout the different layers at Neve David, coupled with a good state of preservation on the one hand but lack of preserved hearths or trampled floors on the other hand, all suggest reduced mobility and a departure from ephemeral occupation at the site, which are evident already in the early layers. If a comparison to the three ethnographic above-mentioned examples is in place, then Neve David should be interpreted as a multi-layered location of occupations, usually in the time-scale of at least weeks, if not longer, per each occupation layer.

Occupation Intensity Trajectories

The flint and bone weights generally increase over time, according to the three sets of levels analysed in this study. However, not all proxies show a linear trajectory (see also Liu et al., 2020, Fig. 12 for a study of the flint assemblage in Trench VI), a phenomenon for which we have no explanation.

Regarding bladelet cores, a main proxy in the current study, their frequencies of bladelet cores are telling: They fluctuate at the bottom part of the sequence, and then decrease towards the upper levels. The somewhat lower frequencies of bladelet cores in the upper levels (compared to the middle levels) may reflect a more ad-hoc approach to bladelet production, reminiscent of ethnographic examples where the more sedentary a group is, the less investment in core preparation is evident (Parry & Kelly, 1987). However, the densities of cores and of bladelet cores increase through time, a trajectory which may indicate higher occupation intensity.

Examining the Neve David sequence, as represented by the three sets of levels, we do not observe significant changes in the spatial patterns of material remains, but the density of finds (lithic and bones) do display a diachronic increase, even though some fluctuations are present (see also Liu et al., 2020). It has been suggested that occupation intensity at GK sites increased due to their function as large aggregation sites leading to reduced mobility (Kaufman, 1992; Liu et al., 2020; Maher, 2019). This in turn is argued to act as a pivotal step paving the way to sedentary life and related demographic, ecological and cultural processes well established during the Natufian (Bar-Yosef & Belfer-Cohen, 1989; Stutz et al., 2009; Valla, 2018; Weissbrod et al., 2017; Yeshurun et al., 2014a, 2014b). Based on our findings we interpret the archaeological deposits at Neve David to represent reduced mobility, already occurring during the earliest phases of occupation at the site. We associate the increase in artefact density as evidence for a gradual increase in group size. However, demographic growth was probably limited since there are no indications for the exhaustion or diversification of local resources at Neve David. The hunting strategy at the site does not differ significantly from earlier Upper and Epipalaeolithic examples, as opposed to the broad spectrum foraging observed at Natufian sites (Bar-Oz et al., 1999; Stutz et al., 2009; Yeshurun et al., 2014b). It is challenging to unequivocally determine whether the increase in site occupation intensity at Neve David is related to gradual and continuous reduction in mobility and development of a higher degree of sedentary lifeways, while group size remained unchanged, or that the observed trajectory resulted from a gradual increase in group size, due to either demographic growth or social aggregation. Still, our suggestion for increased group size will be tested with higher-resolution archaeofaunal studies and other seasonal indicators to identify shifts during the site’s occupation span.

Located on the edge of the coastal plain of the Mediterranean southern Levant, the Neve David case-study presented here is unique in the combined and multi-disciplinary analyses conducted in the search for settlement patterns in the Middle Epipalaeolithic. Comparisons to other excavated or tested Early and Middle Epipalaeolithic sites such as Hefzibah (Ronen & Kaufman, 1976; Ronen et al., 1975) and Nahal Hadera V (Godfrey-Smith et al., 2003) some 35 km as the crow flies, are hampered by the fact that similar analyses were not conducted at these sites. However, in the steppic east of the Levant, a contemporaneous but distinct process took place, as is evident by the mega-sites in east Jordan, such as Kharaneh IV and Jilat 6 (Maher, 2017, 2019; Richter & Maher, 2013; Richter et al., 2013). A comparison of the two areas on the levels provided here are beyond the scope of the current paper.

Conclusions

In this study we aimed to better understand the nature of increased occupation intensity argued for GK sites in the Southern Levant, a notion that commenced in the late 1980s with the first Neve David excavations. We applied an integrated macro- and microscopic analysis of archaeological finds coupled with spatial analysis. This allowed us to better understand the site formation processes, a crucial step in assuring the integrity of the archaeological record at the site and establishing the foundation for an archaeological interpretation of past human activity. Post-depositional disturbances seem to be minimal at Neve David and we therefore suggest that the spatial distribution of the finds mostly represents human use of space. Throughout the thick (> 1 m) sequence of occupation episodes at Neve David the general pattern of space use seems very similar, with the lack of distinct division of space, defined hearths or well preserved trampled floors. Based on ethnoarchaeological reference and comparison with microarchaeological studies at other Palaeolithic sites, we conclude that this pattern probably represents reduced mobility, though earlier similarly well-studied examples are still rare in the Levantine record. Additionally, the density of the finds generally increases within the upper layers of the site which can be interpreted as either a progressive reduction in mobility or limited increase in group size. Although we could not clearly determine the cause for an increase in site occupation intensity during the GK, we do exhibit the usefulness and importance of an integrated macro–micro approach and the importance of studying site formation processes before establishing archaeological interpretation. Lastly, our findings further highlight the position of the GK as a pivotal stage between highly mobile societies during the Upper Palaeolithic and early Epipalaeolithic and the sedentary and demographically-larger Natufian societies during the Late Epipalaeolithic period. More studies are now required to apply a similar approach to other sites and establish whether our findings and interpretation have a wider implication for understanding the changes in mobility patterns during the Epipalaeolithic.