Introduction

The Bell Beaker phenomenon is characterized by a set of typical artefacts that spread throughout Europe and North Africa between ca 2700 and 2000 BC. A so-called “Bell Beaker package” consists of bell-shaped beakers, archery equipment, daggers, and V-perforated buttons (Harrison 1980; Vander Linden 2007; Shennan 1976). The primary vehicle responsible for the spread of stereotyped Bell Beaker material culture is considered to be shared ideology (Brodie 1997; Shennan 1976), marriage networks (Vander Linden 2007), migration (Price et al. 2004), or economic exchange (Childe 1925).

Despite the large-scale expansion of the Bell Beaker phenomenon, there is a tension between the normative Bell Beaker material culture categories and their local objectification in the form of real artefacts. Various studies demonstrated the strong effect of local conditions on the variability of Bell Beaker ceramics (Jorge 2009; Pearson 1995; Rehman et al. 1992; Sarauw 2007; Všianský et al. 2014). Limited attention, however, has been paid to the relationship between the general category of Bell Beaker projectile points and region-specific effects (Mendoza 2016; Nicolas 2013) that shaped the transformation of this mental image into a stone artefact. In Moravia (Czech Republic), the Bell Beaker projectile points were investigated from the perspective of typology, chronology, technology, and archaeological context (Kopacz and Šebela 1998, 1992; Kopacz 2013; Kopacz et al. 2009, 2003; Matějíčková 2009; Olivík 2009), but little is known about the factors affecting shape variation.

Stone projectile points provide an extraordinary opportunity to evaluate how much regional and local factors influenced the general category of such points. Stone itself is widely available and is a technologically effective material for the production of chipped tools in most geographical areas (Ellis 1997; Knecht 1997). These tools carry a critical source of information about past societies in the form of shape variation, which embodies an intersection of cognitive models, technological choices, and environmental constraints. Variation of stone tool shape reflects spatial and temporal trends that provide a basis for modeling long-term processes and spatial fluctuations of diverse human groups (Ahler and Geib 2000; Amick 1995; Bleed 1997; Kuhn 1994).

The understanding of the morphological variability of stone tools stemmed from the Bordes-Binford debate during the 1960s and 1970s. According to Bordes, variation in stone tool morphology reflects group identities of the producers of these tools (Bordes 1961). Binford’s processual approach emphasizes that stone tool variation is primarily functional because performance of different tasks, such as maintenance and extractive tasks, results in different tool assemblages (Binford and Binford 1966). This discussion, focused on the relationship between morphological variability and function, resonates in subsequent works (Dunnell 1978; Jelinek 1976).

Other authors argued that style, as a material representation of group or individual identity, is functional in terms of non-verbal communication (Wiessner 1984, 1983; Wobst 1977) or that style and function can be understood only in terms of the other (Sackett 1982). Despite these issues, stone tool morphology should allow the tool to perform the activity it was designed for (Bleed 1986), including artefacts that were designed primarily for communicative purposes such as Bell Beaker flint daggers (Sarauw 2007). Style preferences and cultural change could be reflected in spatial and temporal variability (Buchanan and Collard 2007; Buchanan and Hamilton 2009; O’Brien and Lyman 2002; O’Brien et al. 2001). It is possible, however, to isolate some basic aspects of functional variability other than style (White 2013). Moreover, ethnographic evidence showed that functions could vary and that projectile points could even be used as a universal tool (Nelson 1997).

Recent debates about the size and shape (i.e., projectile point style) of stone tools take into account multiple factors, including reworking or resharpening (Andrefsky 2006; de Azevedo et al. 2014; Bradbury and Carr 1995; Castiñeira et al. 2011; Shott et al. 2007; Shott and Weedman 2007; Towner and Warburton 1990) and skill (Bamforth 1991; Clark 2003). Univariate models of skill measurement were not sufficient; so, an approach combining experimental and archaeological material (Darmark 2010) and the multivariate approach were developed to measure and quantify symmetry (Ferguson 2008; Hardaker and Dunn 2005). Different authors emphasized that raw material quality and availability have an important influence (Amick 1995; Andrefsky 1994; Flenniken and Raymond 1986; Tankersley 1994); however, experiments did not support the hypothesis that differences in raw materials are the most significant factor explaining size and shape variability (Eren and Roos 2014). These approaches elucidate not only variability among stone artefacts per se, but also the processes that took part during the life cycles of the artefacts.

On the methodological level, analyses of projectile size and shape have shifted from the qualitative description of lithic artefacts and typology procedures (Cauvin 1974; Finkelstein 1937; Gopher et al. 1991) to more quantitative approaches. Univariate morphometric analyses were based on methods developed decades ago for the identification of complex weapons in North American samples of bifacial tools (Hildebrandt and King 2012; Hughes 1998; Shott 1997; Thomas 1978). Various morphometric methods have been used successfully (Erlandson et al. 2014; Hildebrandt and King 2012; Riede 2009; Wilkins et al. 2012; Yaroshevich et al. 2013), some of them focusing on ballistically significant attributes (Hughes 1998; Shea 2006; Sisk and Shea 2011, 2009). Progress in biological morphometrics and computing has enabled researchers to apply geometric morphometrics to projectile shape analysis in recent years (Buchanan and Collard 2010; Cardillo 2009; Iovita 2011). The geometric morphometric approach facilitates the statistical analysis of shape and size variation based on coordinate data and yields easily interpretable numerical and visual results (Buchanan and Collard 2010; González-José and Charlin 2012; Charlin et al. 2014).

The aims of this paper are to explore shape and size variation of Central European Bell Beaker projectile points from Moravia (Czech Republic) and evaluate how much was this general category of Bell Beaker stone artefacts influenced by local and regional factors. Hence, (1) we explore size and shape variability of projectile points to distinguish shape categories and compare artefact attributes with ecological and cultural variables of Bell Beaker communities. More specifically, we look into whether group-specific Bell Beaker point shape categories are (2) spatially localized or distributed randomly in geographic space and (3) whether they are related to the raw material used, and (4) the degree of reutilization. The methodology takes advantage of Procrustes geometric morphometrics, size analysis (centroid size), retouch index analysis, and resampling statistical methods.

Materials and methods

The sample consists of 194 projectile points from 54 Central European Bell Beaker sites (2500–2300/2200 BC) distributed in the Morava River catchment in the Czech Republic (Fig. 1; Supplementary Table 1). Most of the artefacts were obtained from published literature (Kopacz et al. 2009; Kuča and Kazdová 2012; Olivík 2009; Pavelčík 1974), mostly made of a known lithic raw material. The majority of projectile points were found in graves (144 pcs, 74.2 %); less represented are settlement finds (24 pcs, 12.4 %). Rare finds originated from a cave (3 pcs, 1.5 %) and a hoard (1 pc, 0.5 %). The rest of the assemblage consists of surface finds (19 pcs, 9.8 %) and points from unknown contexts (3 pcs, 1.5 %).

Fig. 1
figure 1

Location of Bell Beaker sites in Morava River catchment, Czech Republic

Full size image

The size and shape of the projectile points were studied by landmark-based geometric morphometrics (Dryden and Mardia 1998), which allows quantitative comparisons of shapes based on Procrustes morphometric and multivariate analyses (Dujardin et al. 2010). Similarly to Buchanan and Collard (2010), shape analysis was performed in the following steps: (1) image acquisition; (2) choice and collection of coordinate data, as the 18 landmarks were distributed along the shape boundary according to the same pattern (Fig. 2, coordinate data in Supplementary Table 2); and (3) Generalized Procrustes Analysis (GPA) to register landmark configurations into optimal registration using translation, rotation, and scaling (Supplementary Fig. 1). The landmarks were distributed along the shape boundary according to the following pattern: LM1—tip; LM2, LM3, LM4—regular intervals between LM1 and LM5; LM5—outer edge of the right wing; LM6—mid distance between LM5 and LM7; LM7—the basal edge of the right wing; LM8, LM9—regular intervals between LM7 and LM10; LM11, LM12—regular intervals between LM10 and LM13; LM13—the basal edge of the left wing; LM14—mid distance between LM13 and LM15; LM15—outer edge of the left wing; LM16, LM17, LM18—regular intervals between LM15 and LM1 (tip). The tangent shape coordinates obtained by GPA and centroid size were recorded. Centroid sizes were measured only in 164 projectile points published in Kopacz et al. (2009). The shape variability of the points was visualized by principal component analysis (PCA) computed from a variance-covariance matrix. Principal components provided guidelines for the description of artefacts.

Fig. 2
figure 2

Projectile point with 18 landmarks used in this study

Full size image

Shape categories (SC) of projectile points were derived from tangent shape coordinates analyzed by the UPGMA cluster analysis algorithm using Euclidian distances as a similarity measure. Optimal number of clusters is indicated with a breakpoint (“elbow”) of sharp and gentle decrease in the plot of within group sums of squares vs the number of clusters.

The pattern of spatial distribution of projectile shape categories in geographic space was assessed based on median distance to the centroid site (MDC), and the randomness of this pattern was tested by statistical methods based on resampling techniques. For each shape category, the centroid site was computed as an “average site” by averaging northern and eastern geographic coordinates of those Bell Beaker sites where projectile points of corresponding shape categories were discovered. Median centroid distance was then computed as a median of distances between all sites of corresponding shape category and their centroid. Median was preferred instead of mean distance (cf. Sosna et al. 2008) to avoid the influence of spatial outliers. A low MDC means that projectile points are spatially clustered, while high MDC values indicate that projectile points spread across a wide region. Median centroid distances calculated from original dataset were then compared to the distribution of distances computed from 1000 random iterations to test whether the shape categories are distributed randomly or are clustered. In each random iteration, the shape categories were randomly redistributed among projectile points (absolute geographic coordinates remained unchanged). Because the shape categories were assigned to points by chance, the distribution of 1000 MDC describe the behavior of MDC in samples where shape categories are distributed randomly in geographic space. Thus, MDC values computed from the original dataset that are lower than 95 % of MDC values obtained from random iterations indicate that shape categories are clustered. If MDC values computed from original dataset lie within 95 % of MDC values derived from random iterations, shape categories are distributed randomly over the area of interest.

The raw material of each projectile point was identified by standard petroarchaeological analyses; the results of which were already published; so, the material designation in this paper is based on existing publications (Kopacz et al. 2009; Olivík 2009). The materials come from various sources across Central Europe. The most common material is silicite of glacigene sediments (SGS) that originated from the area of Northern Europe and was naturally transported and accumulated across Poland to the northern borders of the Czech Republic (Přichystal 2013). The other important raw materials are “Krumlovian forest II” (KL II), “Jurrasic flint of Cracow-Czenstochowa” (JFCrCz), and “Krumlovian forest I” (KL I). Whereas JFCrCz was collected or mined from Jurassic sediments in nearby Cracow in Poland (Přichystal 2013), both KLI and KLII originated from neogene chert accumulations redeposited from Jurassic sediments in the area of southern Moravia in the Czech Republic (Přichystal 2013). Other imported materials (radiolarite, chalcedonite) are rare in the research sample. The spatial distribution of lithic raw materials was evaluated by MDC resampling analysis (see above).

The effect of reutilization on Bell Beaker projectile points was studied using the retouch index (RI). Each blade element was divided into eight sections, and every section was evaluated for the presence of retouch flake scars. Retouch is defined as secondary chipping along the edge that is found over the original or previous flake scars. The RI is then computed as a sum of all section scores divided by the total number of sections (Andrefsky 2006). The index expresses the extent to which a biface was resharpened or reconfigured. It is supposed that such resharpening or retouching of the hafted biface is an effort on the part of the tool maker or user to prolong the useful life of the tool. Hence, the RI is a measure of tool curation (Andrefsky 2006). Retouch indexes obtained were compared with Andrefsky’s experimental results, which illustrate different stages of reutilization. The retouch analysis was done only on 32 projectile points from the Hoštice site, the largest site in our assemblage.

Landmarks were collected by CLIC software (Dujardin et al. 2010). Geometric morphometric analysis was carried out in R, using the procGPA function of the “shapes” package (Dryden and Dryden 2012). Resampling analysis was performed in MS Excel 2003 using Visual Basic scripts.

Results

Shape and size of points

All recorded, rotated, and scaled points are graphically summarized in Supplementary Fig. 1. Scree plot suggests that three to six principal components are sufficient to describe the data but the plot does not show a clear bending point (“elbow”) (Supplementary Fig. 2). The final choice of five components is a trade-off between the result of the scree plot, the amount of variability explained (75.2 %), and the ability to interpret the meaning of the components (Fig. 3). PC1 (34.3 % of the variation) reflects length to width ratio, and PC2 (18.3 %) corresponds to shape of base (convex or concave shape). Artefacts’ symmetry is reflected by PC3 (9.7 %). PC 4 (7.6 %) and PC5 (5.3 %) illustrate shape variation of projectiles’ wings.

Fig. 3
figure 3

Shape variation of projectile points along first five principal components (mean configuration ± three standard deviations)

Full size image

Variability of shape indicates the existence of several shape groups of projectile points. The results of UPGMA cluster analysis suggest that there are ten different clusters (shape categories): one dominant cluster, six smaller clusters, and three clusters composed of single artefacts (outliers) (Supplementary Figs. 3 and 4). According to the plot in Supplementary Fig. 2, eight or ten clusters can be distinguished. We inclined to ten-cluster solution because it provides the interpretation more suitable for the issue discussed. The characteristics of shape categories are summarized in Table 1 and Fig. 4.

Table 1 Shape categories of projectile points.
Full size table
Fig. 4
figure 4

Mean configuration of 10 shape categories

Full size image

Shape category 1 (SC1) represents the most common group and comprises 75 % of the assemblage. Points of SC1 are symmetrical, longer, have a concave base and conspicuous wings (Fig. 4 (1)), and are larger in size. Points of SC2 are rather short and wide on average, with a concave base and conspicuous wings (Fig. 4 (2)). They are symmetrical and small in size. Projectile points in SC3 are longer and symmetrical, with no distinctive concave base and no distinctive wings (Fig. 4 (3)), and are moderate in size. Projectile points in SC4 tend to be short and asymmetrical (Fig. 4 (4)), and small in size. SC5 is represented by a single damaged artefact. The distal part of this projectile is lost, and its convex base is reminiscent of SC6 (Fig. 4 (5)). Projectile points in SC6 are characterized by a convex base, notch (Fig. 4 (6)), and a large size. Projectile points in SC7 exhibit a damaged and/or reutilized shape whose base resembles SC1 and 2. The points are broken (Fig. 4 (7)). Projectile points in SC8 are asymmetrical and, therefore, similar to SC4 in shape (Fig. 4 (8)), but rather moderate in size. SC9 and 10 are outliers. SC9 represents artefacts with a characteristically shaped base, significant length (Fig. 4 (9)), and a very large size. SC10 has a specific shape with a notch, prominent wings (Fig. 4 (10)).

Spatial distribution of shape categories

Polygons (convex hulls) superposed on map show spatial distribution of the shape categories in Morava River catchment (Fig. 5). Figure 6 shows the median distances to the centroid for seven shape categories calculated from both original dataset (points) and 95 % of random iterations (boxes). SC5, 9, and 10 were not included in the analysis, as they consist only of a single point. The results indicate that shape categories are distributed randomly over the area of interest, as their MDC computed from original data lies within 95 % of the MDC obtained from iterations with a random spatial pattern (Fig. 6). The only exceptions are shape category 8, whose MDC values lie at the lower limit of 95 % of MDC values produced by random iterations. This indicates that projectile points of SC 8 may be clustered in geographic space.

Fig. 5
figure 5

Polygons (convex hulls) for spatial distribution of ten shape categories in Morava River catchment

Full size image
Fig. 6
figure 6

Comparison of median distances to the centroid (MDC) calculated from original dataset (solid points) and 95 % of random patterns (boxes) in seven shape categories (SC1–4 and SC6–8)

Full size image

Raw material

The distribution of lithic raw materials across shape categories is shown in Table 2. Given the small size in SC3–10, only two major shape categories (SC1 and 2) are further compared. In the first shape category (SC1), SGS predominates among the raw materials (21.8 %), followed by KL II (17.0 %), JFCrCz (14.3 %), and KL I (12.2 %). In SC2, the most common lithic raw material is SGS (33.3 % of artefacts), as in SC1. Other recorded materials for points included in SC2 are KL I (20.0 %), KL II, and JFCrCz (both 13.3 %). The comparison of SC1 and 2 suggests that the distribution of raw materials is similar between the two shape categories (goodness-of-fit, P = 0.87).

Table 2 Lithic raw material by shape categories
Full size table

The results of spatial analysis of the distribution of raw materials in geographic space are presented in Fig. 7. For all raw materials, MDC computed from original sample lies within 95 % of random iterations. This means that despite the extensive geographical variability in the sources of raw materials, the raw materials were randomly distributed among the sites.

Fig. 7
figure 7

Comparison of median distances to the centroid (MDC) calculated from original dataset (solid points) and 95 % of random patterns (boxes) in five lithic raw material categories (KL II, SGS, JFCrCZ, KLI, Radiolarite)

Full size image

Reutilization

Retouch indexes (RI) computed for 32 points from the Hoštice site (Supplementary Table 3) and indexes based on Andrefsky experimental reutilization stages are shown in Fig. 8. Our comparison suggests that RIs for the Hoštice site are similar to those of the first reutilization values of Andrefsky. This means that the points found at the Hoštice site were reutilized, but that the effect of reutilization on the shape of points was negligible.

Fig. 8
figure 8

Distribution of retouch indexes of 32 points from the Hoštice site and Andrefsky experimental reutilization phases (New, Series 1–5)

Full size image

A comparison of PC1 values, centroid size, and RI clearly shows that there is no relationship between the length/width ratio and RI, or even a relationship between size and RIs (Fig. 9). Similarly, shorter shapes of SC2 do not have high values of the retouch index (Supplementary Table 3).

Fig. 9
figure 9

Scatterplot of retouch indexes (RI) of 32 points from the Hoštice site and score on the first principal component (PC1) displayed by the centroid size (CS) of points

Full size image

Discussion

The most important shape attributes of projectile points of Central European Bell Beaker sites are relative length (longer or shorter shapes), shape of base (convex or concave), symmetry (symmetrical or asymmetrical shapes), and shape of wings (distinctive or indistinctive). What does it tell us? Shape of the projectile point represents a variable entity, which could be reduced into several categories of different significance.

There is no straightforward way to interpret the pattern of shape and size and their relationship. Functional aspects such as ballistics, killing power, and the kind of delivery system are often emphasized (Binford and Binford 1966; Dev and Riede 2012; González-José and Charlin 2012; Christenson 1986; Iovita 2011). Simultaneously, it is directly or indirectly assumed that function is associated with style (Nelson 1997; White 2013). Symmetry/asymmetry could illustrate the complexity of factors affecting shape. On the one hand, symmetry/asymmetry could be interpreted either as an expression of skill (Darmark 2010) or as a precondition of desired ballistic qualities such as rotation (Lipo et al. 2012). Such complexity makes the interpretation of shape and size challenging. That is why we classify the projectile points into characteristic shape categories, which allow us to evaluate technological and cultural attributes, their spatial and contextual distribution, the lithic raw material used, and the effect of reutilization.

The classification of projectile points to ten shape categories is based on the most important shape attributes explored by Procrustes shape analysis (relative length, shape of base, symmetry, and shape of wings). The dominance of SC1 (147 out of 194 points) reflects a high degree of unification in the production of projectile points by Bell Beaker people. This distinctive shape probably expresses a regional Bell Beaker cognitive model of how a projectile point should be shaped: longer in shape with distinctive wings and a negative base. SC2 (15 points) and SC4 (seven points) are the categories with only minor differences from the dominant SC1 category. A shorter shape and small size typical for SC2 could relate to reutilization. SC4 represents asymmetrical projectile points, whose shape might be related to reutilization, low skill, or specific ballistic purposes. SC3 (six points) and SC8 (five points) are long, asymmetric, and have a narrow base. The narrow-based shapes, however, represent a rather marginal fraction of the sample. They can, therefore, hardly represent a subject of any remarkable interpretation. SC3 and SC8 do not fit in with other Western European Bell Beaker arrowhead shapes because they lack a notch. SC6 (four points) and SC10 (one point) has a notch typical of Late Holocene pressure flaked arrowheads of Western Europe (Apel 2012). It is another tradition with a different hafting technique. Clearly broken or strongly reutilized shapes are classified as SC5 and 7. Outlying point of a large size (SC9) was recognized by some authors (Kopacz et al. 2003) as a possible point of composite dagger. Moreover, some shape categories could be failed attempts to create an ideal point shape. This failed attempts could be related to raw material quality, availability, knappers skills, or reutilization.

We observed low level of regionalization in the area. The most common shape categories are distributed randomly throughout the region. No regional preferences for specific shape categories were recorded with the possible exception of SC8, which tends to surround the Morava River. This category is characterized by a narrow base and was produced from material that originated in the area north of Carpathians (SGS and JFCrCz). Moravian Bell Beaker sites are situated in lowland areas, which were favorable for Neolithic habitation. We therefore cannot suppose a significant environmental influence in comparison to the other periods of agricultural prehistory.

The results suggest that, at least in the two most numerous shape categories (SC1 and 2), there is no relationship between projectile point shape and raw material the point is made of. The association might exist, however, in SC3 and SC8. These categories with narrow-based points contain no artefact made of Krumlovian forest chert, which was exploited in western Moravia and are exclusively made from material originated from the area north of Carpathians. It can be explained by the fact that points of SC3 and SC8 are spatially located in broad valleys that naturally connect Transdanubia and North European lowlands.

Most of the projectile points were found in graves as part of funeral inclusions. They have to be approached as results of specific mortuary practices that do not have to mirror the reality of everyday life directly. Moreover, the projectile points appear in graves in various locations (Matějíčková and Dvořák 2012). This challenges the view that they were always parts of complete arrows embedded in shafts. Even in living societies, hunters and warriors carry only limited number of precious shafts along with additional arrowheads that can be attached and used in case of necessity. The fact that the majority of projectile points in our sample come from burials reinforces the interpretation that shared ideology of how the world should look like played one of the key roles in the creation of limited variation of projectile point shapes.

The overall shape uniformity of the projectile points and the limited effect of raw material and geographic location have a few consequences for the interpretation. First, high degree of uniformity in shape suggests the existence of shared ideal image of a weapon that accompanies its bearer in the region of our interest. This is in agreement with the understanding of strong Bell Beaker ideology of gender differences and warrior/hunter ethos (Fokkens et al. 2008; Sarauw 2007; Sosna 2012). The lack of association between shape and geographic location as well as shape and raw material indicates high degree of interaction among the communities in the region. Exchange of ideas, gifts, and humans within the region is the most parsimonious explanation for the pattern observed.

A strong interaction between the communities resulted in a low variation of point shapes and none or a weak spatial clustering touching a principal question of Bell Beaker spread across Europe. Although genetic and isotopic studies stress the effect of migration during the third millennium B.C. (Allentoft et al. 2015; Price et al. 2004), there is also an alternative argument that Bell Beaker is neither a culture nor population but rather a phenomenon that resulted from interaction among persons and groups (Vander Linden 2007, 2009, 2016). This does not mean that there was no gene flow and humans did not migrate during the third millennium B.C. in Central Europe but it means that the observed uniformity in Bell Beaker point shapes is unlikely to be the result of a single factor such as migration.

Conclusion

The most important shape attributes of Bell Beaker projectile points found in Central Europe are the ratio between the length and width of points, shape of their base, point symmetry, and shape of wings. These attributes, together with point size, constitute what we call style. Although our research identified several stylistic categories of points, there is a strong degree of uniformity in the Bell Beaker sample. The most dominant stylistic category (SC1) constitutes 75 % of the sample. The dominant stylistic category was pervasive across geographic space and was not significantly affected either by raw material or by reutilization.

The rest of the variability can be described as shape categories of smaller size and only minor shape differences in comparison to dominant group (SC2, SC4), clearly broken or strongly reutilized shapes (SC5 and SC7), uniquely shaped points (SC9, SC10), shape categories with a notch associated with a different hafting technique (SC6) and spatially specific narrow-based shape categories associated with lithic raw materials originated north of Carpathians (SC3 and SC8). The shape categories with a narrow base were distributed closer to the Morava River, which probably served as an axis for the spread of materials, ideas, and humans in Late Neolithic times.

We interpret the lower degree of reutilization of points as a consequence of a non-utilitarian role of projectile points included in Bell Beaker graves. The projectile points found in these graves are possibly artefacts with potentially strong symbolic meaning tied to the identities of the dead.