Stylistic Study of the Late Mesolithic Industries in Western France: Combined Principal Coordinate Analysis and Use-Wear Analysis

3.1 Choice of the Corpus

Several habitats occupied during the Late Mesolithic in the West of France were selected. The corpus includes settlements that underwent archaeological operations (excavations or test pits) with published radiocarbon dates (Table 1). These two conditions ensure the comparability of the collected data. The selected sites are relatively well distributed across the study area, although most of them are located on the coast (Figure 1). In fact, only Pont-Glas (Marchand, Le Goffic, & Marcoux, 2011; Marchand, Goffic, Donnart, Marcoux, & Quesnel, 2017) is located inland; Beg-an-Dorchenn (Rault, 1992), Beg-er-Vil, and La Gilardière (Marchand, 1999) are in coastal situation.

Figure 1

Location of the selected archaeological sites for the study (CAD: L. Hauguel-Bleuven).

The bibliographic data available for each site were consulted and analysed, in order to retain information relating to the archaeological intervention, raw dates, and technological data relating to the lithic industry. It was possible to consult the collections resulting from this work for Beg-er-Vil and Beg-an-Dorchenn, and observations from these sites were particularly useful for the use-wear study. The choices made in the different corpus led to a representation bias between the technological and the use-wear study (Table 2). This difference is less significant for Beg-an-Dorchenn, as the collection could be seen by all the authors (the 12 pieces missing from the typo-technological analysis correspond to armatures that were too fractured to be accurately classified).

3.2 Technological Approach

Technological choices are approached through different characteristics considered to structure the production of blanks: the percussion techniques used, when they could be estimated (Pelegrin, 2000), the original knapped block (core, debris, blade, or flake), the volumetric designs, and orientation of debitage (unipolar, bipolar, discoid, etc.). Subsequently, elements related to armature shaping are retained, such as the methods of blank truncation or the type of retouch used to regularise the edges (Inizan, Reduron, Roche, & Tixier, 1995). This information is noted on a presence/absence basis in a database (Table 3).

3.3 Typological Approach

The general morphology of the armatures was also taken into account for redefined types. Structuring attributes were selected, such as the general shape of the tip (triangular, trapezoidal, or pointed), the symmetry axis of the piece and the delineation of truncations (Inizan et al., 1995; Perrin, 2001). These characteristics have been ranked from the most general to the most specific (Figure 2). Three classes of armatures are thus considered, according to their shape. These classes are then divided into sub-classes, based on the symmetry axis of the piece. Finally, types are defined according to the delineation of each truncation. This classification makes it possible to describe the pieces more accurately and creates a uniform interpretative grid for the studied sites. Morphological information is given in the form of counts (Table 4).

Figure 2

Hierarchisation of the visible characteristics view on microliths, with the different possibilities observed (CAD: L. Hauguel-Bleuven).

3.4 Statistical Approach

Statistical tools are used to compare the typo-morphological assemblages of armatures from each archaeological site, as well as the technical gestures applied to each of them. The aim here is to estimate the differences between the assemblages, in an objective way. In order to do this, a dissimilarity matrix is constructed for each dataset as a first step. Since there are two kinds of datasets, one with counts and one with presence/absence information, two dissimilarity indices are used here. The Jaccard index is used for the technological dataset, which contain qualitative information. Based on the counting of common characteristics, it is the best suited to this type of data (Hammer, 2002). The typo-morphological information is handled with the Bray–Curtis index, because the information is quantitative here. This index compares the population gap of the different objects studied here, and divide them by the total population (Greenacre & Primicerio 2014, pp. 62–63; Hammer, 2002).

Then, these matrices are analysed by using a Principal Coordinate Analysis (PCoA), an ordination method adapted for this kind of information (Carlson, 2017, pp. 297–303; Gower, 1968; Rigaud, 2011, p. 55). PCoA is combined with ascending hierarchical classification (HC) (Shennan, 1988), in order to better visualise the clusters.

The aim of this approach is to not only assess the technical similarity of the occupations, but also the degree of resemblance of the armature’s collections. These two aspects are studied separately, by calculating two PCoAs: one for each dataset. The cross-referencing of these results aims to evaluate the influence of technical gestures on the morphology of the armatures. Statistical processing is carried out in the Rstudio software, version 3.6.2.

3.5 Use-Wear Methodology

In order to question the function of the trapezoidal industries of the Late Mesolithic period in Brittany, a use-wear analysis was carried out. The discovery of several arrow shafts armed with trapezoidal bitruncations (Kozlowski, 2009), preserved in an archaeological context in Northern Europe, led researchers to interpret the geometric industries as projectile elements (Marchand, 1999). Nevertheless, a precise understanding of the functioning of these types of tools can only really be grasped from a use-wear approach (Semenov, 1964). The functional interpretation of arrowheads can not only involve the examination of certain types of diagnostic fractures, but also requires an integrated analysis of all the macro and microscopic attributes of the impact wears, on large samples of materials (Coppe & Rots, 2017; Rots & Plisson, 2014).

In our case, we decided to systematically study and describe all the macroscopic and microscopic attributes of the impact wears of the trapezoidal bitruncations from the sites of Beg-er-Vil (Quiberon, Morbihan) and Beg-an-Dorchenn (Plomeur, Finistère), to be able to compare the functional data with PCoA. Macroscopic attributes were observed, such as the initiation of removals (Ho Ho Committee, 1979), their shape (González Urquijo & Ibáñez Estévez, 1994), their termination (Fischer, Hansen, & Rasmussen, 1984), their length, quantity, orientation, and distribution over the active area of the tool. A total of 381 armatures were studied. Microscopic attributes have been described in terms of micro-topography, brightness, orientation, and distribution (González Urquijo & Ibáñez Estévez, 1994). Several analytical experiments with trapezoidal bitruncations allowed us to establish a use-wear reference collection to infer the function of the archaeological traces (Calvo-Gómez, 2018, 2021).

4.1 Technical Choices

Before the calculation of technical dissimilarity index, data selection was necessary to limit the over-representation of “zeros.” For this, the technical gestures that are absent in all sites studied here were removed from the table. The same treatment is made for techniques thar are site-specific (Table 5). After the computation of the PCoA, the justified variance reaches 80.55% on the two first dimensions. In addition, the correlation between the original distance and the results of the PCoA give a good coefficient, which is equal to 0.98. To put it differently, there is a good correspondence between the results and the original values.

The projection of the two first dimension brings into opposition the Beg-er-Vil levels and the other occupations (Figure 3, left). The HC confirms this observation (Figure 3, right). The stratigraphical unit 101 and 4 of Beg-er-Vil are quite similar, with a medium-low dissimilarity index between 0.25 and 0.50 on a scale from 0 (perfect similarity) to 1 (perfect difference) (Table 5). In the same way, the Beg-an-Dorchenn and Pont-Glas sites share a good similarity, with a distance of 0.38 only (Table 5). Finally, La Gilardière getting closer to Beg-an-Dorchenn (with a distance of 0.41), but remains distant from the other sites (Table 5, Figure 3).

Figure 3

Graphical results of the PCoA (on the left) and its hierarchical clustering (on the right), based on the technological data (CAD: L. Hauguel-Bleuven).

This bipartition seems due to the absence of microburin technique at Beg-er-Vil, but also to the retouch used in the shaping of microliths (Table 3). The microburin technique is a method of blank fragmentation: a notch on the blank is made by percussion and enlarged until this bipartition (Figure 4, 1–3). The fracture formed by this process looks like a trihedral point, sometime visible on the microlith (Figure 4, 4b). Characteristics debris are also produced, called “microburins” (Figure 4, 4a) (Albarello, 1987; Inizan et al., 1995, pp. 84–86; Rozoy, 1968). This technique is identified at Beg-an-Dorchenn, Pont-Glas, and La Gilardière not only by the presence of “microburin,” but also by the conservation of the trihedral point on some arrowhead. Moreover, these sites show a greater diversification of retouch type, compared at Beg-er-Vil, on which only the direct abrupt and the crossed retouched are known (Figure 5a and b). The rest of the reduction process is rather homogenous between the Mesolithic sites studied here. In fact, the direct (hard hammer) and undirect percussion (punch) are used during the exploitation of unidirectional and bidirectional cores, and core on flake. At Beg-er-Vil and Beg-an-Dorchenn, discoïd cores seem to be used for the blank production too (Table 3 and Figure 6).

Figure 4

Explanatory diagram about the “Microburin technique” (1–3) and the reproduction of one microburin (4a) and one trihedral point (4b) (Inizan et al., 1995, p. 85).

Figure 5

Explanatory diagram which shows the direct retouch (a) and the crossed retouch (b) (Inizan et al., 1995, p. 148).

Figure 6

Core plate. Morphology: (1 and 2) unipolar core; (3) multipolar core; (4) bipolar core; (5 and 6) core on flake; (7) discoid core; Source: 1, 2, and 6, La Gilardière (Marchand, 1999, pp. 392 and 394); 3, 4, and 5, Beg-er-Vil (Marchand, 1999, pp. 419–420); 7, Beg-an-Dorchenn (Rault, 1992, p. 46).

4.2 Morphological Choices

As for the technical choices, a selection of the information was necessary to limit the over-representation of “zeros.” The types which are site-specific and whose representation is lower than 4% of the total population are removed before the calculation of the dissimilarity matrix. This threshold is chosen arbitrarily to reduce the number of “zeros” without depleting too much dataset. Moreover, the level 101 of Beg-er-Vil is also excluded, due to its low number of microliths (Table 4). After the computation of the PCoA, the justified variance reaches 79.93% on the two first dimensions. In addition, the correlation analysis between the original distance and the results of the PCoA show that there is a good correspondence between them (with a coefficient equal to 0.96).

The projection of the PCoA show the opposition between Beg-er-Vil and the other sites, along the first dimension (Figure 7, left part). La Gilardière seems to be isolated in the second dimension too, and is located in the superior right corner of the diagram (Figure 7, left part). Pont-Glas and Beg-an-Dorchenn share the lower dissimilarity index, with a distance of 0.52 from each-other (Table 6). In a general way, the sites are quite unsimilar with indices between 0.50 and 0.80 (on a scale from 0 to 1). The HC reflects this despite the absence of clear bipartition (Figure 7, right part).

Figure 7

Graphical results of the PCoA (on the left) and its hierarchical clustering (on the right), based on the typological data (CAD: L. Hauguel-Bleuven).

Beg-er-Vil shows a high number of symmetrical trapezes with straight truncations and a great number of symmetrical trapezes with concave truncations (Figure 8, numbers 1 and 4). Beg-er-Vil is also marked by the poor representation of points, more frequent at Pont-Glas and Beg-an-Dorchenn (Table 4). La Gilardière is marked by a higher quantity of asymmetrical trapezes with concave truncations as well as by the presence of symmetrical triangles with straight truncations (Table 4 and Figure 8, number 15). Moreover, this site is also characterised by the presence of some “armatures à éperon,” specifics at La Gilardière (Figure 8, number 19). Finally, Pont-Glas and Beg-an-Dorchenn share some microlith types, more precisely among the points: points with oblique truncation, points with retouched based, and backed points (Table 4 and Figure 8, numbers 16, 17, 18, 20, 21, and 22).

Figure 8

Selection of Microlith. Typology: (1 and 2) symmetrical trapezes with straight truncations; (3–5) symmetrical trapezes with concaves truncations; (6 and 10) asymmetrical trapezes with concave truncations; (7–9) asymmetrical trapezes with straight truncations; (11–13), asymmetrical triangle with straight truncations; (14) Chatelet point; (15) symmetrical triangle with straight truncations; (16 and 17) truncated point; (18), point with retouched base; (19) “pointe à éperon”; (20–22) backed point. Source: 1, 4, 11, Beg-er-Vil (Marchand, 1999, pp. 427); 2, 3, 6, 7, 8, 9, 10, 12, 14, 17, 18, La Gilardière (Marchand, 1999, pp. 399–400); 5, 13, 22, Beg-an-Dorchenn (Rault, 1992, pp. 73–76); 6, 7, 16, 17, 20, 21, Pont-Glas (Marchand et al., 2017, pp. 262–263).

4.3 Results of the Use-Wear Study

The study of impact wears provided numerous results concerning the function and functionality of Teviecian bitruncations. On one hand, 11 out of the 62 trapezoidal armatures from the Beg-an-Dorchenn site were interpreted as impacted armature elements. Generally speaking, the asymmetrical bitruncations bear impact wears parallel to the large base of the trapeze, i.e. they were impacted as piercing arrow elements (in 32.9% of cases; Figure 9a and b). On the other hand, symmetrical bitruncations often bear impact traces organised perpendicular to the large base of the trapeze, suggesting the use of transverse-edged arrows (in 18.5% of cases; Figure 9e). These impact wears consist of mainly fractures and scars of the active areas, especially on the large base of trapezes, as well as microscopic linear elements that follow the direction of the macroscopic removals. Most of the time, functional removals are fractures of the angles, with very different shapes. The fractures are usually caused by hard materials, that creates cone initiations, and have feather or step terminations. Finally, the rare microscopic attributes related to impacts accompany the direction of removals in the form of striations with a rough bottom.

Figure 9

Impact wears of the Beg-an-Dorchenn armatures. (a) Burin-like fracture (×5). (b) Bending initiated scars (×10). (c) Isolated striation, parallel to the axis of the trapezoidal bitruncation (×500). (d) Burin-like apical fracture (×8). (e) Scars and fractures on the large base of a symmetrical bitruncation (×5).

On the other hand, the use-wear study of the 319 bitruncations from the Beg-er-Vil site allowed us to identify 49 elements with clear impact wears. Almost systematically, the symmetrical bitruncations were also interpreted as transverse arrowheads, due to the location and organisation of the macro- and microscopic attributes, which are always perpendicular to the large base of the trapeze (Figure 10). The initiation of the removals associated with a projectile impact was predominantly conical, with very pronounced and glossy striations. This suggests impacts against hard mineral materials. The predominant morphologies were elongated or pseudo-burin type removals starting from the corners of the large trapeze base. Again, in the studied Beg-er-Vil collection, removal terminations have often feather or step shapes.

Figure 10

Impact wears of the Beg-er-Vil armatures. (a) Several scars and fractures on the large base of a symmetrical trapeze, with marked cone initiations (×5). (b) Burin-like fracture (×10). (c) Microscopic striations, perpendicular to the large base of the symmetrical trapeze (×100). (d) Several impact scars on the edge of a symmetrical trapeze (×5). (e) Cone-initiated scar, with a semi-circular shape and feather-like termination (×10). (f) Bright microscopical striations, running perpendicular to the large base of the trapeze (×100). (g) Burin-like fracture (×10). (h) Burin-like fracture (×10). (i) Scars and fractures with cone initiations (×10). (j) Microscopic linear traces, running perpendicular to the large base of the trapeze (×100).

5.1 The Technical and Typological Variations in Regard of the Chrono-Cultural Framework

Several ethno-archaeological studies highlight the relationship between the morphology of the produced objects and technical choices. This is a recurrent research topic in works on traditional pottery craftsmen. It appears that certain technical aspects carry strong identity markers, transmitted through generations of craftsmen, while others attest to direct contacts (at markets or during visits) or indirect contacts (through the diffusion of finished products) between potter communities (Gosselain, 1992, 1995, 2017; Roux et al., 2017). These works highlight the social and significant character of technical choices for the populations that implement them. In our study, recurrent technical gestures could point to a common basis for knapping and shaping lithic tools. This is the case for percussion techniques (direct and/or indirect) and the volumetric organisation of cores (mainly unipolar frontal debitage with one or two successive debitage planes), identified in all the examined occupations. On the other hand, other technical elements, such as microburin technique or retouch types, are more likely to be local choices (Beaune de, 2008; Leroi-Gourhan, 1973; Roux, 2015). This mixture of technical, aesthetic, and social characteristics makes it possible to define “communities of practices.” This expression refers to the behaviours shared on a daily basis by a human community. The term “practice” refers to the gestures and techniques used as well as to the objects used and the relationships between the actors (Wenger, 2000). Keeping this in mind, the PCoA of technical information would therefore indicate the coexistence of three communities of practices in Western France during the Late Mesolithic (Figure 3).

These three communities of practices echo the chrono-cultural framework known in this region. The Late Mesolithic of Western France is characterised by the opposition between Tevecian and Retzian, defined by Rozoy (1978, pp. 747–748). A bipartition of Tevecian group has been proposed by Marchand, who identifies two successive phases, the Beg-er-Vil facies and the Hoedic facies (Marchand, 2014). Furthermore, the spatial distribution of symmetrical trapezium subtypes in Brittany points to three different regions in the sixth millennium cal BC. (Marchand, 2000, 2005). The definition of these groups is made on the basis of the microliths typology. Tevecian sites are marked by high proportions of trapezoidal bitruncations (symmetrical or not) and asymmetrical triangles (Figure 8, numbers 1–13), while the Retzian group is defined by a more diversified quiver and the presence of “pointes à éperon” or Chaville points (Figure 8, numbers 14 and 19) (Marchand, 2000, 2005; 2014, pp. 234–239; Rozoy, 1978, pp. 747–748). Nevertheless, the distinction between Tevecian and Retzian sites is not as clear as we thought. In fact, La Gilardière get close to Pont-Glas and Beg-an-Dorchenn on the basis of their dissimilarity index (Table 5). From a technical point of view, Tevecian and Retzian debitage are quite similar with the exploitation of unipolar core by indirect or direct percussion (Marchand, 2014, pp. 234–239). The absence of microburin technique is specific to the Tevecian. But this observation can be restricted to the “oriental Tevecian,” since only Beg-er-Vil stand out by this aspect.

5.2 The Function of Microliths

Although the statistical results show that there is no strict correspondence between technical gestures and the final shape of the armatures, their functioning distinctly relies on their morphology. Use-wear analyses provide useful hints to objectify the functioning of the Mesolithic arrowheads. Even though there are always some exceptions to the rule, impact wears show clear tendencies on the fitting of the arrowheads, depending on their geometrical shape. Symmetrical trapezes almost always have impact wears running perpendicular to the large base of the trapeze, meaning their use as transversal arrowheads. On the other hand, asymmetrical trapezes and triangles deliver frequently impact wears on the apical angles, parallel to the axis of the armature, which indicates their use as piercing arrowheads.

Thus, we consider that the statement that the morphology is not essentially linked to their use (Pfaffenberger, 1992) must be, at the very least, relativised. It is clear that the function of the archaeological tools cannot be exclusively inferred by their shape alone. Since their very beginning (Keeley, 1980; Plisson, 1985; Semenov, 1964), use-wear analyses have repeatedly demonstrated the spectrum of potential technical choices while the object implemented are much wider than their morphology would initially suggest. However, in our opinion, while these observations are objectifiable, they also imply the meaninglessness of the shape of an object. Since the shape of an object is an independent character of their function and their social meaning, the shaping of an object is meaningless.

We consider on the opposite that the morphology could be understood as an actant role on the potentialities of the object. In our case, the strict correlation between the use of symmetrical trapezes as transversal arrowheads is a good example. Their shape seems to be a defining character of their use. The same statement is valid for the asymmetrical bitruncations. While the human techniques can make several technical procedures with a similar tool, the tool itself is an actant. Social meaning of the objects is not determined by their morphology, but by their use. The social meaning of the symmetrical bitruncations is totally meaningless without their use. The social meaning of the hunting activities and their techniques themselves is, in our opinion, what fulfils the identarian role of these objects.

5.3 What About the Concavity of Some Truncation?

By studying the technical and functional aspect of microliths, one last thing can be highlighted here: the variation in the delineation of the retouched line used to shape the armatures. If a straight line of retouch is enough to determine triangle from trapezoidal bitruncations or to differentiate the symmetrical and asymmetrical one, some truncations are marked by a deeply concave retouch. This fact is all the more striking because it is less visible, or not at all visible on the object once hafted. Moreover, the delineation does not seem to have a function: this characteristic can help to join the microlith and the shaft on the microliths, but it is not mandatory. In fact, symmetrical trapezes with straight truncations are hafted in the same way that similar pieces with concave truncations. Furthermore, this characteristic seems to be specific to some places. For example, the assemblage of Beg-an-Dorchenn is known for the important presence of trapezes with concave truncations, while Beg-er-Vil is marked by the predominance of straight bitruncations. The delineation of the truncation can be another characteristic of style, like the retouch type (Lenoir, 1975). But the discretion of this feature puts it closer to the notion of “adjunct form,” an element not necessary for the use of the tool but present for other reasons, aesthetical of identical ones (Sackett, 1982, p. 71).

5.4 Limitations of This Study

Through this study, the difference between Beg-er-Vil and the rest of the corpus has been highlighted, on a technical and typological point of view. This observation can be due to a real distinction, or to the constitution of the corpus itself. In fact, only four archaeological sites have been chosen here, corresponding to occupations which have been studied and properly dated with radiocarbon. But this selection limits the number of comparisons possible. Moreover, the technological and typological information are uneven between different sites. Actually, the data come from former and published studies for Pont-Glas and La Gilardière. More recent publications have been consulted for Beg-an-Dorchenn and Beg-er-Vil, and the re-examination of the lithic assemblage has been possible too. In consequence, Beg-er-Vil has more complete information about its assemblage, which is probably reflected in the statistical analysis presented here.

This fact constitutes the most important limit of this study. A small corpus makes the difference between the number of sites and characteristics more important, which can affect the results of the analysis and his reliability. About this point, it seems that the PCoA presented here are quite reliable, with great goodness-of-fit. Second, this lack of information makes impossible the use of PCoA, or other statistical analysis, to compare the functional information between sites. Effectively, this kind of data concern only two sites, which make useless this method.