Waste Around Longhouses: Taphonomy on LBK Settlement in Hlízov

2.1.1 Hlízov – Small LBK Settlement

The site of Hlízov in eastern Central Bohemia was chosen as a case study. During systematic research in 1982–1986, a small, early LBK settlement consisting of three long houses and a group of 26 sunken features was uncovered (Pavlů, 2002, pp. 45–116). In addition to Neolithic objects, several Bronze Age features were found during the research. However, these more recent components were sufficiently distant from the Neolithic, so the LBK settlement at Hlízov can be considered relatively isolated and undisturbed by more recent human activity. It is for this reason that this site has been chosen as an ideal place for the research of the above questions. From a regional perspective, Hlízov is located in a warm fertile area that has been intensively inhabited since the Neolithic. Four kilometres to the north of Hlízov, the Elbe River, the largest river in Bohemia, flows and has been an important communication corridor since Prehistory.

In total, 34 features were found during the survey that can be associated with the LBK settlement. It is possible to identify 3 longhouse plans and 25 pits of various sizes and shapes. Most of the features found were within Area A (Figure 1), but two Neolithic features were isolated from it, 160 m to the east (Area B).

Figure 1

Plan of excavated areas. Upper-right corner – map of the Czech Republic with the position of the site (GMRT – Ryan et al., 2009).

None of the houses found can be said with certainty to have been excavated entirely (Figure 2). While houses 12 and 14 had preserved post holes and external ditches, a typical feature of the architecture of the early LBK, house 13 had only three central rows of columns. It could be a relic of an unfinished house, and the archaeological record shows only a phase of construction without external walls and pits for clay extraction. The presence of a possibly younger house between two older ones would correspond for example to the similar situation in the nearby Miskovice settlement (Last, 1998; Pavlů, 1998). It is also possible that it was an atypical building, lacking surrounding structures such as lateral pits. However, we cannot comment further on the house in either case due to its unusual nature and the absence of representative artefact assemblages.

Figure 2

Map of Area A of the Hlízov site where Houses 12, 13, and 14 have been almost fully excavated and pit 32 probably belonging to another house, named House 35. The dashed line shows the hypothetical extent of the pits in the Neolithic period.

It is assumed that there was another house to the east of the excavated area (No. 35). Of this one, only the south-western part, consisting of a piece of a trench and a side pit (No. 32), was probably uncovered (Pavlů, 2002, p. 47). Although pit 32 has not been excavated entirely, the greater part of it has been explored. Despite the presumed presence of other pits around house 35, their assemblages were not excavated. For this reason, the rich assemblage from Pit 32 is included in this study as the only representation of a material with a close spatial relationship to house 35. However, a more detailed characterization of this house will be avoided for this reason.

The presence of other houses and significant concentrations of settlement pits in the vicinity is unlikely. During the excavation, survey trenches were excavated with an orientation that should have revealed other similar finds.

In addition to the houses, 24 pits have been excavated, most of them in the vicinity of houses 12 and 14. These pits vary in shape, size, and depth; however, they all have bowl-shaped and irregular bottoms, and none were regularly shaped like storage structures (Šumberová, 1996, p. 67). Although 24 features were recorded, they may have looked different in the Neolithic period. Due to the erosion of the original terrain level on the site (that is common for the majority of LBK settlements – Květina & Hrnčíř, 2013, p. 326), some of the features must have originally been bigger and thus connected into larger pits. Today, we can only see the bottoms of some of them (Figure 2 – dashed line). Thus, the groups of pits were probably originally larger sunken units that ran along the sides of the two houses. Instead of numerous smaller pits, the features can be imagined as long lateral pits, typical of LBK. The most noticeable are lateral pits 9 and 11 for house number 14, pits 10 and 16 for house 12, and pit 32 for house 35. These pits were excavated in sectors and mechanical layers during the fieldwork, allowing the structure of their fills to be studied.

2.1.2 Artefact Assemblage

Pottery and lithic assemblages were examined. The occurrence of other artefacts and ecofacts on the site, such as polished stone tools, ground stones, or supply materials (unworked non-local stones), was also observed.

The collection of Neolithic pottery found at the settlement consists of 1088 fragments, of which 563 specimens can be identified. Twenty-one sunken pits containing pottery were examined across the site. However, only 8 of these contained a representative assemblage that would be usable in statistical analyses (i.e. 15 or more fragments).

The lithic industry found at the settlement consists of 143 artefacts representing a weight of 275 g. The material was uncovered from 8 pits and two pieces are coming from the surface. Almost all the material was discovered in one pit (n = 120). The other structures provide less than 8 stones.

In addition, 2 polished axes, 6 abraders, 2 grinding tools, 23 fragments of ground stones (abraders/grinding tools), 10 PRTs (pebble rock tools), and 60 non-local stones for possible tool production (weighing a total of 10.5 kg) were found at the settlement.

The bone assemblage consisted of approximately 50 fragments. However, the bones were very poorly preserved and did not allow detailed osteological analysis. Despite this, 10 bones of large herbivores (mostly cows) were selected from the assemblage and sent for C14 dating. Due to the poor preservation of the collagen in the bones, the C14 dating did not provide any usable data.

2.2.1 Pottery Analysis

2.2.1.2 Fragmentation

For individual sherds, the degree of fragmentation was observed. The tracking of fragmentation assumes that a highly fragmented assemblage has undergone multiple transformations or destructive events that have gradually broken the originally large sherds (Kuna, 2015, p. 282). Thus, the greater the fragmentation of a ceramic set, the longer the taphonomic path of the artefacts can be assumed (Vondrovský, 2021, p. 67).

To record the degree of fragmentation, indexes are conventionally used that attempt to compensate for different fragmentation tendencies in different vessels. The first to be used was the SW (size/wall) index, expressed as the ratio of the longest measurable dimension of the sherd to the thickness of its wall (Květina, 2005). However, this index is misleading for sherds with significantly thin or thick walls. Therefore, an alternative index was created, the Fragmentation Index (IF), which, unlike the SW index, works with the weight and thickness values of the shard and compensates for the index’s tendency to distort extreme values of wall thickness (Kuna & Němcová, 2012, p. 185; Pilař, 2022, p. 61). Although the IF was able to resolve the original SW index distortion and is very useful for example, for simulating the gradual fragmentation of assemblages (see Kuna & Němcová, 2012, p. 189), we do not consider it the optimal way to characterise fragmentation. This is because it ignores the value of fragment length, which strongly influences fragment tendency to break (Gifford-Gonzalez et al., 1985), and in some cases, it gives the same value to a rounded and a long, narrow shard, which has different tendencies to break up. It was therefore decided to create a new index, which would again work with the length and thickness of the ceramic shard but would eliminate the possible tendency of the index to correlate with extreme values of wall thickness.

The calculation of the new Length Fragmentation Index (LFI) is similar to the SW index, except that the wall strength value is dynamically adjusted for different values. For the SW index, extremely thin sherds had very high values and, conversely, thick-walled ceramics had low values. This phenomenon was compensated for by the unlinear lowering of the wall thickness value. The resulting formula is as follows:

LFI = S W 0.345 / 25.5 .

This formula can easily be used for example as Excel expression = (S/W ^0.345)/25.5. The wall (W) square root value was modelled on a set of 16,000 ceramic fragments from the Neolithic period for which the necessary metric values were given. The ceramic collections come from the sites of Prague – Krč, Bylany, Hlízov and Statenice.^[1] No correlation was found between the resulting values and the wall thickness of the fragments when using the above formula (Table 1). Dividing the resulting value by 25.5 was chosen to average the LFI value to 1. The LFI is thus a convenient way to express the fragmentation of a shard using a single number.

Table 1

Table of the dependence of the fragmentation indexes (SW index, IF, LFI) on the ceramic wall thickness

The data used come from four Neolithic sites at which this relationship was observed.

2.2.1.3 Refitting

Refitting – the dispersal of fragments from individual vessels – has been observed for the entire ceramic assemblage. The aim of this method is to trace the links between the fragments and to see where the ceramic assemblage may have been dispersed by different processes (see above). This approach thus introduces a dynamic element into the observation of an otherwise static archaeological situation (Cziesla, 1990, p. 17).

The study of refitting has a long tradition in archaeology. It is typically applied to bones, stone tools or ceramics and is used, for example, to determine the separation of stratified layers (Morin et al., 2005; Plutniak, 2021) or more generally to characterise a situation taphonomically (Kuna & Němcová, 2012, p. 197; Květina & Končelová, 2011b, p. 63). The scattering of fragments has also been interpreted in some cases as the result of symbolic behaviour – for example, social enchainment, in which fragments were actively moved within and between sites (Chapman, 2000, p. 138). Refitting testimony varies from material to material – in the case of chipped industry, production sequences are traced, and refitting illustrates the transfer of waste/products from production (Cziesla, 1990, p. 15). In contrast, the refitting of pottery and bones reflects the dispersal that occurred only at the end of their primary function (Morin et al., 2005; Plutniak, 2021). Although it is possible to look for direct connections between the pottery and bone fragments, they cannot be assembled into meaningful sequences as in the case of the chipped stone industry.

In the case of pottery refitting, several categories of association were used, based on the type of association or its probability. These categories are based on the work of Bollong (1994, p. 18), but have been simplified for the purposes of this article (Figure 3):

1. Direct connection of two sherds.

2. Similarity of two sherds based on their similar wall thickness, colour, ceramic texture, curvature and decoration.

3. Similarity of two sherds based on their equal wall thickness, colour, ceramic texture, curvature, but without decoration.

4. Sherd without probable refitage – “orphan” sherd.

Figure 3

Refitting Categories. On the left are the connection/similarity categories used by Plutniak (2021). On the right are the three categories used in the Hlízov database and in the middle is a graphical representation of them.

Of the categories used, only the first is “connection” (Plutniak, 2021, pp. 5–6). However, categories at the level of “similarity” must also be used – the set studied was of poor quality and heavily abraded, so most of the more distant refits were made up of categories 2 and 3. Nevertheless, it would be a mistake to overlook this relationship, even if it is limited by its empirical nature.

Refitting can be evaluated in a number of ways, but graphical representation remains essential as mathematical representations are not always fully informative (Plutniak, 2021, p. 4). Although different ways of presenting data can change the meaning of the data (Bollong, 1994, p. 18), there is no ideal visualisation that can be applied universally. Therefore, combining several seems to be a reasonable approach.

A common practice in LBK settlement research is the use of simple diagrams (classic examples are e.g. Allard et al., 2013; Stäuble, 1997, p. 84). These diagrams are a simple but clear way of expressing relationships, which are then interpreted empirically. However, this approach becomes unusable with large amounts of data. A good alternative is therefore the use of network analysis, where individual sherds/components are converted into nodes and their links into edges (Knappett, 2013). Network analysis methods make it possible to track the location of individual nodes in the network, identify clusters, bridges or isolated groups. Thus, network analysis provides a fast and efficient method of exploring relationships that is applicable to large volumes of data. The disadvantage is that it is less readable for spatial data.

This study thus used the plotting of data at the level of individual spatial units (division by objects, layers and their sectors, Supplementary 2) into a Fruchterman–Reingold plot to analyse the underlying data distribution and connectivity. Subsequently, the connections were plotted in a Geographic information system (GIS) to understand the real spatial distribution of connections.

Several approaches can be used to quantitatively characterise contexts based on refitting data. In general, we can describe the levels of internal cohesion of a spatial component or the degree of admixture with other components. The TSAR method was recently developed to detail the cohesion and admixture between two spatial units (Plutniak, 2021). However, it requires highly detailed data collection, such as recording directly adjacent sherds, making it impractical for large-scale assessments. To simplify the process, a method using the ratio of intra- and inter-joints to characterise cohesion was employed. Unlike the “mixing” value of Morin et al. (2005), which used the ratio of inter-layer refitted fragments to all refits in a layer, this study used the percentage of intra vertices (or “self-loops”) within each spatial unit. The resulting cohesion ratio reaches 100% when all refits are within the same spatial unit and drops below 50% when most refits connect fragments from different units.

However, the cohesion ratio can be distorted in some cases. An object divided into many smaller units is likely to have lower values than an object with a few large sectors. At the same time, this value can vary considerably for spatial units that are poorly represented. However, it is useful for comparing units of similar size, or for general characterisation of cohesion and mixing.

2.2.2 Study Method of Lithic Industry

To evaluate the composition of the lithic industry, we questioned the representativity of the productions chaînes opératoires (operational chain – hereafter CO) on the site. This approach is based on technological analysis of the artefacts that aims at determining the intentions of the knappers and the gestures made to reach these intentions (Pelegrin, 1995). This is possible by reading diacritical schemas on the artefacts (order and sense of the removals negatives) to determine the place of the studied artefacts in the different sequences of the production. In the best cases, physical refitting enables one to get a detailed description of the operative chains. In the absence of such possibility, mental refitting offers a less precise (Pelegrin, 1995) but still relevant picture, to discuss the general organisation of the production and the presence or absence of stages of the CO at a location.

This work is of course conducted after sorting the raw materials. They are frequently treated differently during the Neolithic (Binder & Perlès, 1990). The identification of the raw material has been conducted by the naked eye. Raw materials are very well studied in the Czech Republic and surrounding regions. All the characteristics of the different outcrops documented in the area are described and documented at different scales by Přichystal (2013). The macroscopic classification has been reviewed by Miroslav Popelka – Institute of Archaeology, Charles University. Sometimes, for specific raw materials, their characteristics are so specific that it is possible to work at another level, one of blocks or macro-features groups (Denis, 2019). This allows the possibility to go beyond the refitting by highlighting links between the artefacts.

We compared the composition of the lithics through quantitative and qualitative data. The quantity is measured by the number and weight of artefacts to evaluate the density per pit (see above).

The typological classification allows us to assess the nature of discarded artefacts, and we evaluate the representativity of the CO by constructing techno-economic diagrams (Geneste, 1985; Perlès, 2016) when the artefacts are numerous enough to be statistically representative. Otherwise, data were treated simply by presence or absence.

2.2.3 Statistics and Software

Excel, Power Query, and Access were used for data collection and manipulation. Further, Jamovi, SPSS, and R software were used for statistical and network analysis. QGIS was used for spatial analysis and visualisation.

In addition to descriptive methods, statistical testing was performed. For this purpose, ANOVA with Tukey’s post hoc comparison method was used (Ostertagova & Ostertag, 2013, p. 258). The conditions for these tests – normality and homogeneity of data were checked using Kolmogorov–Smirnov and Shapiro–Wilk normality tests and Levene’s homogeneity test. If the conditions were not met, the sample size was checked (Ostertagova & Ostertag, 2013, pp. 258–259).

Next, network analysis was used and was performed in R with the library “igraph.” The Fruchterman–Reingold layout was used (Hevey, 2018, p. 310).

Factor analysis was used for the final synthesis of the results, for which varimax rotation was used (Shrestha, 2021, p. 7). Clusters of features with similar factor loadings were then visualised in space using QGIS.

3.1.1 Pottery Fragmentation

Overall, the assemblage can be characterised as above-average fragmentation. The average LFI value of the Hlízov pottery is 0.8, which is considerably lower than the value of the Neolithic settlements on which the index was calculated (the average was set at 1.0 – check Section 2.2.1.2). At the same time, it is more fragmented than the pottery from Bylany, a densely built-up settlement, where the LFI was 0.98 (Pilař and Květina, 2023). However, this value may not necessarily be the result of different waste management, but could, for example, be caused by a lower quality of local pottery.

Fragmentation varied considerably from feature to feature (Figure 4 and Supplementary 3). Differences are also visible in individual parts of the fill. However, there is no clear spatial pattern in the settlement between pits with high and low fragmentation rates.

Figure 4

Fragmentation of pottery in individual pits. Boxplot visualisation.

3.1.2 Density

The density of pottery fragments is highly variable for individual features (Supplementary 4). Some pits were practically free of ceramics (density up to 5 frag/m³), but others reached up to 20 times these values (83 frag/m³). The pottery densities do not create a clear pattern in space. However, it is interesting to note that both houses 12 and 14 have one lateral pit with an extremely high concentration of pottery (Figure 5). The presence of similar “dominant pits” with a high concentration of waste has already been observed in Miskovice or Bylany (Last, 1998, p. 26; Pavlů et al., 1986, p. 307).

Figure 5

Pottery density in individual pits visualised by bars.

3.1.3 Vessel Shape, Decoration and Its Technique

According to relative chronology, the set of pottery corresponds to the transition of early and middle LBK in Bohemia. Elements of early LBK are, for example, the presence of wide engraved lines or the frequent occurrence of thin-walled bowls instead of globular pots (Neustupný, 1956; Pavlů & Zápotocká, 2007, p. 29). Among the elements associated with middle LBK are the presence of notenkopf punctures (present only on one sherd), decoration with thin engraved lines, or a higher proportion of thin-walled pottery (Pavlů, 2002, p. 63). Evidence of the early LBK influence can be observed also in the more archaic architecture of the houses.

Although the fragmentation of the pottery does not allow a clear reconstruction of the decorative motifs, the frequent presence of two parallel lines meeting in a V-shape, from which they diverge into two arcs, indicates the presence of the typical “A” style. This motif on the bowls is considered typical of the beginning of the Middle LBK (Jíra, 1910, pp. 74–75; Pavlů & Zápotocká, 2007, p. 31; Soudský, 1954, pp. 89–91).

The main shapes of the vessels in the assemblage can be characterised as follows (Figure 6, Table 2):

Figure 6

Main vessel types representation.

Thin-walled globular-shaped vessels made up 15% of the assemblage and were mostly decorated with engraved curvilinear decoration. Thick-walled coarse pots made up 10% of the assemblage and had knobs and occasional technical decoration (nail marks). Thin-walled bowls were the most common, accounting for 54% of the assemblage. 40% were decorated with engraving. Both recti and curvilinear lines occurred, but they could be part of the same motif (“A” style). Less common were thick-walled bowls (8%, knobs, engraved decoration), bottles (4%, engraved decoration), and storage jars (8%, knobs, handles). The prevalence of bowls contrasts sharply with, for example, the Bylany assemblage, which represents the majority of the Middle and Late LBK, where bowls make up only 20% of the assemblage (Pavlů, 2000, p. 141).

When focusing on the decorative elements and their technique (Květina & Končelová, 2011, p. 200), the ensemble appears homogeneous; 98% of all Hlízov linear decoration is made up of engraved lines only. This complete uniformity of the decoration creates a contrast with most LBK sites in Bohemia, where the chronology is based on the relative proportion of linear decorations (Pavlů et al., 1986, pp. 340–352).

But the engraved lines differ. According to the descriptive system, we distinguish between “grooves” – wide lines typical for early LBK and classic thin lines that disappear only in the final LBK (Pavlů et al., 1986, pp. 340–352; Vencl, 1961, p. 102). But the boundary between these styles is unclear, and its identification varies from person to person. Therefore, a more detailed approach has been taken, based on the metric and morphological characteristics of the engraved line (width and cross-section).

Differences in the technique of decoration proved to be distinctive, and the use of thin engraved lines, executed with U- and V-shaped tools, and wider grooves executed only with U-shaped tools can be observed on the settlement. This indicates the use of two different toolkits, probably by two different communities/learning networks. The empirically defined categories of groove × line (gamma × delta in the descriptive system; Pavlů et al., 1986) overlap slightly in line width, but approximately 2 mm can be given as the dividing line.

The differences are particularly noticeable when examining the technique in individual pits. Grooves appeared exclusively in pits 16, 32, and partly 30, while most of the representative pits 3, 9, 10, 11, and 22 contained mostly thin lines (Figure 7). Features 32 and 16 thus represent a separate group in terms of their decoration technique, and pit 30 contains a combination of both technical implementations (Figure 9).

Figure 7

Width of engraved lines on pottery decoration. Boxplot visualisation.

3.1.4 Refitting

A graph built at the spatial unit level creates a discontinuous network (Figure 8). However, most of the nodes are connected to one main network, but 4 nodes are completely isolated and another three, creating their own small networks.

Figure 8

Pottery refitting network between individual excavated units. Fruchterman–Reingold visualisation was used.

In the centre of the main network lie the units of feature 9, which also have the highest degree of centrality. Around it are nodes of other features, but do not create a completely connected network – many units are not connected to the rest of the network directly, but through bridges. Feature 16 is thus connected to the main network indirectly through feature 30. Features 10 and 11 are slightly more connected to the network centre, but their units form a looser, less connected network. It is important to note that in these pits only the upper layers are connected to the centre of the network.

When this refitting network is transferred to the map (Figure 9), one can see that the pits in the settlement are connected to each other. Although the connections might be expected to respect the house unit, a significant number of edges connect distant pits “belonging” to different houses.

Figure 9

Pottery refitting network visualised on settlement plan. Pit 9 is loosely connected to pit 22 (Area B – not shown in picture). Colour groups show two decoration techniques (Section 3.1.3). All visualised intra-feature connections correspond to refit category 2 (similarity of shred and decoration, without physical connection; Section 2.2.1.3).

However, as mentioned earlier, the joints do not form a completely connected network, and on closer inspection, we can observe the disconnected nature of pits 32 and 16, which both lie to the east of house 12. Together with pit 30, which forms a bridge with the main hub of the network, pit 9, they form an interesting group that absolutely overlaps with the results of the decoration technique analysis (Section 3.1.3). Pits 32 and 16 thus form a separate group which has its own distinctive decoration technique and which is not connected to the pits to the west. A mixture of materials from both groups was deposited in pit 30. While house 14 is surrounded by style 1, house 12 is on the border of two groups. House 32 cannot be characterised in this way because it is represented by a single pit (no. 32).

On the basis of refitting and cohesion level (Supplementary 5), we can characterise different structures of the pit infill. The most connected is the fill of pit 9, which forms a complete network – all its units are connected to all the others (Figure 10 – right). A low level of cohesion (below 35%) accompanies this phenomenon. This is the result of a process in which an already highly mixed pottery assemblage entered the pit. Moreover, in the context of the size of the pit, this was clearly not a one-off event, but a trend. Filling in pits 10, 11, or 16, where the network is not fully connected, gives a different picture (Figure 10 – left). The connections between the units may be the result of several phases of filling or of filling with several unmixed assemblages. Thus, the units forming “bridges” may be the place where two natural (e.g. stratigraphic) layers are mechanically connected. Higher levels of cohesion are also associated with this pattern (mean values spread between 45 and 71%).

Figure 10

Pottery refitting network between individual excavated units. Left – lightly interconnected feature. Refits usually connect only neighbouring units (Feature 10). Right – a highly interconnected pit. All units of the object are connected to each other (Feature 9).

Other pits cannot be well characterised in this way because they contain small quantities of pottery or have few spatial units. But generally, their fills tend to be interconnected. There was no situation at the site where the layers were strictly separated from each other, which would be a clear indication of multiple phases of infill/infill by different assemblages (similar to Pilař & Květina, 2023, p. 31).

3.2.1 Quantitative Analysis of the Material

The lithic industry is mostly concentrated in pit 32 (n = 120), which is the south-western pit of House 35 (Figure 11). The pits surrounding houses 14 (pits 3 and 9) and 12 (pit 16) are not nearly as rich since they have respectively delivered 9 and 6 lithic artefacts. Two artefacts have been discovered in pit 23 which is related to a younger cluster of postholes from the Bronze Age in area B.

Figure 11

Repartition of the lithic industry on the site. House 35 = pit 32; House 14 = pits 3 and 9; House 12 = pit 16 (Figure 2).

This quantitative distribution was compared with the volume of the pits to assess the density of lithic material (Table 3).

Table 3

Evaluation of the density of lithic artefacts per pits based on their volume

The table demonstrates that the quantity of lithics discarded is not related to the volume of the pits. Pit 32 undeniably has a concentration of waste. It contains many more artefacts than pit 9 that is bigger or pit 16 that has about the same volume. There could be a difference when comparing the number or the weight of the artefacts. This is connected with the nature of the artefact discovered. Cores are heavier than small chips but both artefacts could have the same meaning in terms of representativity of local production. To better understand waste management, qualitative observations must indeed be considered.

3.2.2 Refitting and Raw Materials Similarities

Refitting has been tested on this assemblage because the complementarity or proximity between the artefacts seemed very favourable (Figure 12). Despite this impression, no evident physical refitting could be carried out.

Figure 12

General overview of the Skršín quartzite artefacts from pit 32. Photography: D. Pilař.

The identification of the raw materials exploited in Hlízov can be used to refine the understanding of waste management. Three main raw materials have been identified on the site (Figure 13): (1) the Skršín quartzite; (2) the Jurassic-Cracow flint and (3) the Erratic silicites (SGS). This area of Bohemia lacks good quality outcrops of siliceous raw materials. The closest sources could be found between 88 and 120 km with the belt of erratic silicites. But we can’t be sure of the exact location of the procurement of this raw material. The most dominant raw material in Hlízov is the Skršín quartzite that can be found 120 km from the site. Finally, Polish Cracow flint outcrops are distant (300 km) from the sites. These distances are considerable and used to qualify exogenous raw materials regarding stone tools productions (e.g. Mateiciucová, 2008; Allard & Denis, 2021 for the LBK).

Figure 13

Map localising the different sources of raw materials exploited in Hlízov based on Přichystal, 2013. Number 1 in blue: Skršín quartzite outcrops; Number 2 in green: Erratic silicite potentialities; Number 3 Jurassic-Cracow flint outcrops (GMRT – Ryan et al., 2009).

It remains quite difficult from a statistical point of view to compare the different features due to the low number of artefacts (Table 1 in Supplementary 6). The Skršín quartzite is largely dominant in the assemblage (more than 75%, n = 108), but this is especially true for the pit 32 (83%, n = 100). Jurassic Cracow flint and SGS have been identified in almost all the pits in small quantities (respectively, 18 and 11 pieces in total). Pit 3 has only delivered one artefact in Cracow flint. The pit 16 is the only one on the site to have delivered burnt artefacts (n = 3). This classification was analysed a bit deeper by identifying some groups among the distinct individualised raw materials, through specific macro-features characteristics. However, this was not possible for each of the raw materials, except Skršín quartzite and the Jurassic Cracow artefacts. Within these two materials, it was possible to distinguish different facies according to the characteristics of the raw material (colour, texture, and inclusions). Without presuming the value attributed to these facies (same block and same micro-deposit), we formed sub-groups of raw material that we used to carry out a network analysis (Figure 14) in order to better understand whether similarities could be identified between the different pits.

Figure 14

Macro-features matching method and network analysis highlighting similarities between the different pits and the degree of cohesion in and between the pits. The number of chipped stones is displayed for each pit.

This parameter would suggest a very high cohesion of pit 32 with lots of similarities of the materials inside this pit. It also demonstrates some rare similarities between pits and more specifically between pits 32, 9 and pit 21. Pit 16 gathers all the burnt artefacts identified on the site, underlying a specific pattern regarding waste management. However, it hides potential relations with other pits because macro-features can no longer be identified after burning.

3.2.3 Typo-Technological Observations

We will now specify how the Hlízov lithic waste is composed, regarding some typo-technological observations. Two elements will be used in this perspective:

1. The toolkit composition (Table 2 in Supplementary 6).

2. The representativity of the different stages of the operational chain.

Houses are formally assigned according to pit position.

Table 4 counts the number of tools and the waste from the tools identified in the collection. Only 16 tools have been identified which is a very low number (11.5%), especially as we counted the pieces bearing macroscopic scars interpreted as resulting from their use. A use-wear analysis would be required to confirm this attribution. The composition of the toolkit is surprisingly not very varied. One scraper has been identified in pit 32. Two sickle blades belong to pit 9 and pit 32. The other pieces are non-very specified retouched pieces. No tool has been identified in pit 16 that could again underline its specificity regarding the other houses.

It is now possible to state which stages of the production are represented on the site. It was previously stated that the raw materials can almost all be considered as exogenous but some sequences of the debitage took place locally.

Most of the artefacts, when their technical characteristics are relevant, are attributed to a blade or bladelet chaînes opératoires (simplified database in Supplementary 7, column 10). This can be diagnosed thanks to the diacritical schemas and the determination of the percussion techniques. Most of the debitage is conducted by indirect percussion. It appears frequent that the cores are then reused primarily as tools (hammerstones for example) or knapped to produce a few flakes that can be retouched. Nine techno-economic categories were created that allow identification of which of the sequences are represented on the site:

1. Flakes from the shaping-out;

2. Flakes from the crest preparation;

3. Flakes from the maintenance of the core;

4. A specific category designated for the rejuvenation of the platform;

5. Blades;

6. With isolation of blades that have transversal negatives from the crest preparation or a natural or cortical facet;

7. With isolation of blades that are more dedicated to correct an accident or help to obtain better convexities;

8. All the pieces that do not have enough specific characteristics to be more precisely attributed to a stage of the production;

9. Cores and flakes that have been reused.

Studies by raw material and by pit were used to identify specific waste management (Supplementary 8).

All the blade CO in Skršín quartzite is represented in pit 32 (Figure 15), except the very first flakes of the shaping-out of the blocks. Blocks could have been introduced on the site marginally preformed. The comparison between the number of artefacts and their weight allows the highlighting of whether the representation of the CO is almost complete, there is very little waste, except for the reuse of the cores attributed to the end of the CO. The flakes are on average less than 0,6 g (orange lines in Table 1, Supplementary 8).

Figure 15

Techno-economic diagram of Skršín quartzite based on the different stages of the blade production identified in pit 32.

In pit 9, two artefacts are in Skršín quartzite but only one can be attributed with certainty to blade production, the second one is less clear but may also be (Table 2 in Supplementary 8). Both flakes are knapped by direct hard hammerstone. One is a big flake of 44 mm in length, 42 mm in width and 10 mm thick that can be attributed to the shaping-out of a block. The second one has 25% natural surfaces and could also be related to this stage. Not only did the characteristics of the raw material bring the flakes from pit 9 closer to a block or micro-facies from pit 32 (Figure 14), but a form of complementarity also appeared between the two pits. The missing stage from pit 32 is represented in pit 9, demonstrating a form of spatial segmentation of the waste in relation to the stages of the blades CO. In addition, the nature of the massive flake is quite different from the contents of pit 32, which is characterised by small elements. Despite the size of the blank, this flake is rough and was not selected as a tool blank.

One flake has been identified in pit 16. It is a very small one (less than 0.1 g), knapped by indirect percussion but it can’t be precisely attributed to a stage of the operational chain. However, by its very nature, it is very close to the elements identified in pit 32.

Only 18 pieces have been discovered of Jurassic Cracow flint (Table 3 in Supplementary Material 8). The paucity of this corpus means that we have to be very cautious when dealing with quantitative data. It can at least be appreciated through the weight of the material (ca. 69 g in total). From a quantitative point of view, Table 5 demonstrates that, as for pit 9 for the Skršín quartzite, pit 3 delivered many more artefacts than all the other pits. This is particularly true for pit 3 where the only artefact discovered is a core (Figure 16, number 1), whose micro-facies have nothing to do with the material from the other pits. Slight differences appear when comparing the different pits (Table 3 in Supplementary 8). Pit 16 only contains a small fragment of the bladelet. Pit 3, as just mentioned, only contains material from the end of the CO. Pits 9 and 32 are quite comparable and pit 21, located in the southern cluster, is the only pit that contains a flake related to the shaping out of a block. So, either the blocks are introduced in the process of debitage with already knapped flakes, or there is a spatial segmentation of the production. The small number of pieces does not allow any clear indication in favour of either hypothesis.

Figure 16

Small blade cores. Number 1, Jurassic Cracow flint discovered in pit 3. Number 2, SGS discovered in pit 32.

SGS raw material exhibits a different pattern compared to the other material regarding the weight of the artefacts (Table 5). The largest pieces are preserved in pits 32 and 16, particularly due to the presence of a core in pit 32 (Figure 16, number 2). Cores could be also introduced in the process of debitage (Table 4 in Supplementary 8). However, for this raw material, the shaping-out of the blocks is not very sophisticated, so this could also explain the absence of this stage on the site. Some flakes with 20–30% natural surfaces or cortex are identified. A form of complementarity between pits occurred regarding this raw material. Indeed, blades from the full sequence of debitage are only identified in pit 9 and are missing in pit 32. In pit 16, the SGS artefact identified is not related to blade production but to the exploitation of a small block as splintered pieces.

Chronological Model

One possibility is that the individual houses and pits were not contemporaneous. For example, we can imagine that houses 35 and 12 were built first, using some of the pits in the area. When these pits were filled in, house 14 was built and used simultaneously with house 12. This would explain the refitting of pottery from the pits near these houses. This model would be consistent with the current idea of the chronology of the pottery decoration, which shows the pottery from pits 32, 16, and 30 to be older. The complementary of the chipped stones from pits 9 and 32 could, according to this model, be explained by the longer taphonomic trajectory of the finds from pit 9. However, this more traditional model is not the only possible explanation.

Cultural Model

Another possibility is that the houses on the settlement were built at the same time and that the observed variability is the result of two communities living side by side – or rather two coexistent traditions reflected in the material culture. The two clusters that occur in the pottery refitting may therefore be the result of different pottery production practices and different waste disposal sites that occurred at the same time. The complementarity of the chipped stone production chain of the two houses would then indicate the interconnectedness of the two groups in the processing of stone tools. In other words, lithic production was centralised at the village level rather than at the household level as suggested elsewhere (the hierarchical structures suggested by de Grooth, 2013; Kegler-Graiewski & Zimmermann, 2003). The idea of diversity at the same stage has also been suggested in other sites (Gomart, 2014, pp. 126–133; Hachem & Hamon, 2014).

Behavioural Model

The behavioural model is the last one we will mention. In this model, the observed variability could be simply the result of different behaviours – the treatment of waste. For example, if differently decorated vessels were perceived differently, they might also have been deposited in other parts of the settlement. This pattern of behaviour could then create two clusters around the contemporary houses, similar to those observed in the case of Hlízov. Regarding stone tool patterns, the variability observed between the different raw materials could also be explained using this model.

In addition to these models, there are certainly many other possibilities that could explain the results. At the present stage of our understanding, we are not in a position to select the most likely of these.

It is crucial to discuss the variability of the data across the settlement and to examine the networks that connect individual houses. If we were to assume spatially defined house units in the first place, we would miss a lot of important data and the results mentioned above would not be identified at all. Using this approach, although we are not able to characterise individual houses in detail (as Hachem & Hamon, 2014; Pavlů, 2000 do), we are able to observe the whole settlement better and more critically. Here we see that it is perhaps better to step back from individual houses and households and look at the site as a whole. Otherwise, we are left with a situation where we can’t see the settlement for the houses.