To answer the question of whether knapping activities had taken place in the excavated area and to reconstruct potential artefact displacements (Bertran et al., 2010), the bearing and plunge of selected artefacts were recorded according to the methods proposed by McPherron (McPherron, 2005 and McPherron, 2018). Moreover, the aforementioned studies have been supported by analyses of spatial distributions of refitted groups of artefacts (Hofman, 1986 and Morrow, 1996). To process and interpret the collected data, open source computer programs PAST and OpenStereo, version 0.1.2 (http://www.igc.usp.br/index.php?id=391), were used (Grohmann and Campanha, 2010). To determine whether the number of finds was reduced by geological processes, the metric structure of the artefacts was compared to experimental data (Bertran et al., 2012, Byers et al., 2015, Dibble et al., 1997, Newcomer, 1971 and Schick, 1986). Particularly important is the percentage of the smallest elements, such as chips, which are the most susceptible to displacement under natural conditions. The lithic material presented in this paper has also been analysed for origin, typology, and technology. It is worth mentioning that lithic refitting studies were a significant contribution to the full understanding of the production of shaped tools.

Microscopic studies of lithic artefacts were conducted as well to determine post-depositional alterations, technological features concerning the type of hammers used during knapping and the function of the specimens. Until now, 72 specimens from trenches 1/2012 and 2/2012, as well as individual finds from the surface of the site, have been analysed. In the course of the study, a metallographic microscope was used (Nikon LV150), adapted for use under reflected light. The artefacts were cleaned to avoid contamination using a warm-water bath containing a detergent and were washed with acetone. The detailed identification of individual traces – micro-scars, striations, and polishes – was performed with magnifications of 50×, 100× and 200×. The interpretation of the obtained microscopic images of selected artefacts was conducted based on comparisons with the results of the experimental tests performed earlier by one of the authors (KP), as well as with those of studies conducted by other researchers. The analysis of single bone remains was performed using taphonomic and systematic criteria.

For the detailed analysis of refit sequences of shaped tools, three refits were scanned using a Diondo D3 CT-scanner. The scans with a resolution of 50 μm were then accurately refitted by hand with the help of the AVIZO and Geomagic softwares. 3D refitting has the advantage of analysing each step of the tool shaping sequence in detail without causing damage to the artefacts by gluing them together and manual handling. It has many further applications, like accurate illustration of each detachment step, analysis of gaps, as well as accurate measurements of flake or nodule metrics and angles. This method was developed recently and was successfully applied to the analysis of stone artefacts (Delpiano and Peresani, 2017, Delpiano et al., 2017 and Weiss, 2015).

Geological studies were based on archival data and observations collected during field survey and excavation. In the framework of the field project, sediment analyses were conducted. To determine the age of the artefacts of the Pietraszyn 49a site, numerical dating methods were used. The absence of well-preserved organic materials at 49a prompted us to use the OSL method.

The first attempt was based on a quartz-luminescence dating experiment conducted in 2014. However, these first OSL dates were contradicting the geological model assumed for this part of Głubczyce Plateau (Jary, 1996) and the expected age range of Middle Palaeolithic occupation in this part of CE (Bobak et al., 2013 and Neruda and Nerudová, 2013). Thus, we decided to take a second series of samples for luminescence dating, now applying the pIRIR225-approach (Buylaert et al., 2009 and Thomsen et al., 2008) to coarse potassium feldspar grains. The methodology of both types of OSL dating techniques is described in Appendix B.

At site Pietraszyn 49a, four sedimentary units were identified, each with different sedimentary environments. They were labelled as units A–D (starting from the top of the section), while the associated layers forming them were marked with Arabic numbers (Fig. 2 and Fig. 3).

Unit D was represented by fluvio-glacial, cross-bedded deposits. The layer complex consisted of sands with an admixture of gravel (up to several %) and clastic sediments (up to 10%; layer D1-D2). In the top part of the fluvio-glacial sediments, iron compounds were accumulated, related to the paleo-soil development processes. We suppose that they could be dislocated clasts of palaeosoils.

The subsequent units C and B consisted of fluvial sediments. Unit C was composed of sand mixed with silt. Within it, there were structures of convoluted deformations. The uppermost part of the layer was eroded.

Unit B was composed of gravel, sand, and silt. On the surface of the eroded units C and D, a layer of sandy silt was formed (B9). This layer forming the basal part of the accumulation of unit B was 10 to 20 cm in thickness. It consisted of a silt fraction of up to 20%. In the top and middle parts of layer B9, the majority of lithic artefacts and single bones were recorded. Based on this finding, it could be assumed that the accumulation of traces of late Middle Palaeolithic human activity occurred after the sedimentation of this layer. The overlying layers of unit B indicated the development of solifluction (layer B8a and B8b) and fluvial processes in small channels (layer B7b). At the bottom of layer B7b, an erosive pavement, consisting of sand and gravel up to 25%, which towards the top layer changed into medium-grained sand with an admixture of clastic sediments up to several percentage points, was observed.

The complex of unit A at Pietraszyn 49a consisted of a modern humus level (20 cm in thickness) and the remnant of features of the Lusatian culture (Bronze/Iron Age).

A small number of artefacts were also found in trench 2/2012. However, the stratigraphy in this part of the site was different than in trench 1/2012. In the bottom of trench 2/2012, fluvial deposits were recorded. It seems that this layer might be an equivalent of unit D recorded in trench 1/2012. The fluvial deposits were cut by a layer of sand and gravel formed during the process of solifluction. The latter layer provided artefacts. Above, loess was deposited, then covered with contemporary soil. We assume that the loess corresponded to MIS 2. It seems that all of the artefacts recovered from trench 2/2012 were moved from higher parts of the valley slopes.

The two dating methods provided divergent results. The previous age estimates obtained using quartz OSL dating suggested that the basement (D) and the lower part of unit B were formed at the end of MIS 3, while the middle part was formed after the LGM (MIS 1). More precisely, layer B9, containing lithic artefacts, yielded numerical dates spanning from 33.9 ± 3.4 (UJK-OSL-49) to 38.7 ± 3.9 ka (KIE-846). The age of unit D was almost equal to that of B9: 33.6 ± 3.3 (UJK-OSL-50) and 39.1 ± 4.0 (KIE-847). Layer B7a was dated back to 13.5 ± 2.0 (UJK-OSL-48) and 14.0 ± 2.1 (KIE-845).

Feldspar luminescence dating (for more details, see Appendix B.2.2) yielded significantly older age estimates. Figure B2 (Appendix B) shows the distribution of equivalent doses of samples L-Eva 1533, 1534, 1535 and 1536, respectively. It seems that the samples from the Troja valley infill were well bleached, whereas the samples taken from below the channel showed a skewed equivalent dose distribution pointing to incomplete bleaching.

The luminescence ages point to a period of fluvial erosion and subsequent aggradation at approximately 60 ka. This finding matches the transition from MIS 4 to MIS 3 (see Fig. 3C, D), from a cold stadial to a more temperate interstadial. All three luminescence age estimates obtained from inside the valley infill yielded equal results, as they overlap within the given error bars (Table 1).

The sample obtained from below the fluvial sequence yielded an age correlated with the penultimate (Saalian 2: Warthe) glacial cycle. Nevertheless, due to the skewed equivalent dose distribution of the sample (Fig. B2: 4), the age might be overestimated and should be regarded as the maximum age.

In conclusion, the dating of the sediments using the pIRIR225-approach is regarded as the most reliable inference concerning diachronic changes of geology, dynamics of site formation processes and chronological frameworks of human settlement represented by an assemblage of the CEM.

In light of the geological analysis and absolute dating, it is possible to present the following scenario focusing on the southern part of the Troja River valley. It seems that the periglacial valley (C and B) was carved into the gravel sand sediments of the Warthe glaciation (unit D), dated to 145 ± 14 ka BP (MIS 6a/MIS 6b). On the slope of the valley above the erosional discontinuity, there were sandy gravel sediments (D1–D2) represented by structures of solifluction lobes, containing the relics of palaeosoils. The age of this soil is unknown: it might have originated from Eem or early Weichselian. Additionally, the age of unit C is not clear. It likely represents one of the first stages of valley development oriented to the north. Based on OSL dating (pIRIR225-approach), it can be stated that the next stage of the shaping occurred at the turn of the Pleniglacial and the Interpleniglacial (MIS 4/3: unit B, 64–60 ka BP). From the beginning of this stage, deposition of stratified sediments occurred with a significant admixture of fine-grained sand. The structure and texture of the lowermost layer B9 indicate that, during accumulation, a significant role was played by the processes of surface flow and downwash, which are regarded as typical for the highest parts of slopes and the initial parts of the valleys. After the deposition of layer B9 in the area of the Troja River valley, human occupation occurred. Afterwards, within the valley, the solifluction process occurred, as documented by layers B8a and B8a. Even later, the valley was cut due to erosion expressed by sedimentary discontinuity and filling of the erosional channel with a straight sequence (decreased grain-size distribution towards the top). Based on OSL dating (L-Eva 1534 and L-Eva 1535), this stage can be placed at the beginning of MIS 3 (∼ 58–60 kyr). During MIS 2 on the slopes of the Troja River valley, loess sediments were deposited. Loess was gradually moved towards the valley bottom.

The majority of the archaeological material was excavated in trench 1/2012 within layer B9 (Fig. 3B–D). A small number of lithics, which could be interpreted as fragments of the same assemblage, were recorded within unit A. They were found in the Bronze/Iron Age feature, which cuts the layers of unit B, in slope deposits of trench 2/2012 and in modern agricultural soil (Table 2). In the following section, we focus on the findings from trench 1/2012.

The majority of finds from trench 1/2012 are represented by lithic artefacts. Only a few bone remains of vertebrates were recorded. The find distribution was limited to layer B9, with a thickness of up to 20 cm (Fig. 4). An exception was the edge of the river bank, where some of the artefacts occurred outside the range of layer B9. It is likely that they were displaced due to downslope movements. These dislocation processes might not have been very dynamic, as the individual artefacts excavated from unit A form refits with artefacts from layer B9.

The density of artefacts can be described as very high (approximately 610 pieces per square metre), corresponding to “high-density patches” (Roebroeks et al., 1997 and Verpoorte et al., 2016). The artefacts were unevenly distributed over the excavated surface. We noted that the finds were clustered in the middle and northern parts of the trench (91.8% of the objects; Fig. 5). At the western edge of the channel, three distinct concentrations (diameters 0.7–0.9 m) were recorded.

To answer the question of whether the finds from the main trench (1/2012), notwithstanding the presence of small clusters, could have been displaced by geological processes, we examined the bearing and plunge of artefacts recorded on the slope and at the flat surface of the riverbank. The Rayleigh test for artefacts bearing on the slope (Z = 3.718; P = 0.024) and the flat surface (Z = 4.436; P = 0.012) indicated a lack of uniformity (Fig. 4 and Fig. 5). In other words, there is a reason to conclude that the orientation of lithics was governed by unidirectional force. The fabric of the analysed samples from the slope and flat surface of the riverbank, according to a Woodcock diagram, appears to be linear (Lenoble and Bertran, 2004, p. 459). The rose diagram indicated that the artefacts deposited on the slope had a tentative orientation along the north–south axis, i.e. parallel to the axis of the valley and corresponding to the decline of the slope of the Troja valley (Fig. 5). This finding could indicate the movement of artefacts located within the slope of the valley.

In the next step of examination, we analysed the orientation of lines connecting particular refits of artefacts (208 vectors). The rose diagram (0–180°) (Fig. 5c) clearly indicated the following trend: •

a large share of the refits were oriented parallel to the axis of the river bank;

This trend indicates that the lithics could have been moved slightly in the northern direction due to the movement of low-energy currents towards the axis of the valley and, at the same time, because of gravity in accordance to slope of the riverbank, i.e. to the east. It cannot be excluded that some larger elements, transferred over a significant distance, are the effects of intentional displacement by humans (e.g., an attempt to clean the surface). Further excavations might help answer these questions. The distances between the refitted elements are 0.1 and 3.4 m, with an average of 1.08 m. These parameters indicate that the displacement of the artefacts might be negligible.

It seems that the remnants of clusters associated with human workplaces could be preserved only in the edge zone of the river bank (Fig. 5b). Three small clusters contained a great number of small finds as chips, suggesting that they could have been accumulated when knappers worked near the ground, kneeling, squatting or sitting (Kvamme, 1997 and Newcomer and Sieveking, 1980, pp. 349–350; Schick, 1986, pp. 36–41). It is also interesting to mention that the largest number of refits with more than three parts came from these clusters. Thereby, most of the refits were recorded in the central and northern clusters (C2 and C3: Fig. 7b and c). Less frequent were refits covering all three clusters, clusters 1 and 2, or only cluster 1.

The results presented above are in accordance with the results of microscopic examinations of lithics. Almost all of the artefacts analysed showed surface alterations. The main post-depositional modification is surface gloss, the intensity of which varies from light to very intense. These changes are visible both with the naked eye and under the microscope. The origin of this modification is usually related to mechanical and chemical processes. Water and sand activities are believed to be the main source of this type of surface alteration (Levi Sala, 1993, pp. 31–32, 71; Plisson and Mauger, 1988, p. 4; Van Gijn, 1990, p. 53). In the case of lithics from Pietraszyn 49a, there was a relationship between the changes noted on their surfaces and the geological context of their deposition.

In conclusion, the unearthed finds from the main trench (1/2012) indicate the presence of remains of some clusters resulting from human activity. Data on the spread of refits and the relief of the valley surface might, however, suggest that these clusters were partially reduced due to the development of slope processes.

Within the top part of layer B9, singular animal remains, such as fragments of horse molars (Equus sp.), were recorded. Unfortunately, they underwent considerable diagenesis due to contact with water (Lyman, 1994, p. 422). It should be noted that these animal remains cannot be directly connected to the late Middle Palaeolithic artefacts, similar to the other open-air sites known from Poland (Wiśniewski et al., 2009). Until now, only at Zwoleń the relationship between human activity and horse remains is documented (Gautier, 2005).

The new chronological data presented in this study question the early ages of finds from the nearby site Pietraszyn 49. These first thermoluminescence dates obtained from sediments yielded ages of 193 ± 17 ka and 130 ± 10 ka (Fajer et al., 2001a, p. 206). Consequently, it was assumed that the inventory from Pietraszyn 49 represents the oldest Micoquian site in central Europe dated back to MIS 6 (Fajer et al., 2001a and Fajer et al., 2001b). However, our studies indicate that artefacts from site Pietraszyn 49 are related to the same time span as lithics from Pietraszyn 49a (MIS 4/MIS 3). The association of both groups of finds is supported by stone tools from site 49, being analogous regarding the state of preservation, raw material, and technology of manufacture, with striking morphological similarities to artefacts from 49a (Wiśniewski et al., 2015). Additionally, they were found at the same altitude (approx. 218 m above sea level) in the same part of the valley.

Only the pIRIR225 luminescence age (about 140 ka) from the basal part of the sedimentary sequence exposed at Pietraszyn 49a, underlying the main archaeological find layer, is comparable to the previous TL ages from Pietraszyn 49. Hence, the high TL ages from Pietraszyn 49 might be explained by contamination of the dated sedimentary units by middle Pleistocene slope deposits. It is important to note, that at site Pietraszyn 49a, almost all of the artefacts (ca. 10k pieces) were found in the layer dated to MIS4/MIS3 (see Section 6.2).

Nevertheless, the new proposal of age assessment of Pietraszyn 49 and 49a does not reject the hypothesis regarding the appearance of asymmetric bifacial tools at the end of the middle Pleistocene in central Europe (Kozłowski, 2014 and Kozłowski, 2016). Their presence has been confirmed within the layers of MIS 7 and MIS 6 at the Biśnik Cave site (Cyrek et al., 2014).

Studies from Pietraszyn 49a deepened our understanding of the process of tool shaping and operational chain economy. Firstly, Pietraszyn 49a shows the phenomenon of individualised treatment of raw materials in the early stages linked to the choice of a particular techno-functional unit sensu E. Boëda (2001), such as the back, playing the role of the prehensile part (Jöris, 2001, Jöris, 2006 and Veil et al., 1994). We noticed that, in some cases, the isolation of the back was integrated with the function of a striking platform, which was used for the detachment of large and flat flakes from the opposite side of the designed working edge of a tool. This idea is well expressed by group Ptr-4. However, these rules were not followed consistently. If there were no conditions for the organisation of such a striking platform, knappers decided to start shaping other parts of the block, for example, from the most distal zones. Secondly, the data also shed light on the question of flexible reaction to some unexpected technical problems caused by unpredictable damage. This behaviour is demonstrated by group Ptr-1 (Fig. 10).

Finally, in the case of Pietraszyn 49a, we are dealing with the technological plano-convex cross-section concept. Refits enabled to reconstruct the sequence of shaping operations. It seems that giving the shape to a flat side was an initial stage of shaping. Only thereafter, the convex side was shaped (e.g., Ptr-1 group). Such a way of tool shaping is in line with the idea of renewing, which is the principle for the concept of an instrument used outside of workshops (Iovita, 2014 and Uthmeier, 2016).

It should be noted that the use of the plano-convex strategy frequently relates to the basal and middle parts of the tools, while for the tips and distal posterior edge, a biconvex strategy was applied (Fig. 8: 3). At this time, we can only speculate on the reasons behind using mixed strategies. However, based on our observations, we cannot exclude that the use of mixed strategies was connected with attempts to adapt to a particular block shape or even the personal style of tool shaping.

Technological organisations cannot be fully characterised without the assessment of scale of tool production. The number of tools potentially produced in Pietraszyn 49a can be assessed based on the study of Acheulean bifaces from Red Barns (Wenban-Smith et al., 2000, p. 239) and Micoquian bifaces from Kabazi V/Crimea, layer III (Veselsky, 2008). For the Acheulian, the average blank (≥ 20 mm)-to-biface ratio is 60:1, while the average blank (≥ 30 mm)-to-biface ratio is 51:1 for the East European Micoquian. Hypothetically, a blank (≥ 20 mm)-to-biface ratio of 38:1 can be assumed for the Pietraszyn 49a site in this context.

The next issue concerns the direction of production. Many CEM assemblages can be characterised by two lines of manufacture, i.e. a core reduction for flake/blade tools and tool shaping based on nodules or chunks and flakes (Boëda, 1997, Boëda et al., 1990, Bourguignon et al., 2006, Frick, 2016, Richter, 1997 and Weißmüller, 1995). In Pietraszyn 49a, the flake/blade blanks of tools were obtained during tool shaping. This hypothesis is confirmed by the presence of a refitted group containing flakes and sidescrapers. As noted above, in Pietraszyn 49a, we did not record any core. This feature makes the complex from Pietraszyn 49a similar to the eastern Micoquian sites in Crimea (Demidenko, 2015 and Uthmeier, 2016), as well as individual sites from central Europe (Mańka et al., 2006 and Veil et al., 1994). This option seems to be a constant element of the operational chain of tool shaping in the CEM. From the perspective of ethnoarchaeological studies, the use of expedient flakes during stone tool production is not surprising. There have been many examples showing that such blanks served to solve “some technical problems” (McCall, 2012, p. 174).

Although discoveries from the Pietraszyn 49a site have been partially moved by slope processes, it is reasonable to question the number of individuals involved in the tool production process. Considering the large number of finds, as well as nodules or chunks, used during manufacture, it seems that we are dealing with remnants of more than one person. Interestingly, the range of technical activities related to the production of end products was quite uniform. They are represented by shaped tools “equipped” with one or two working edges, a point, a back, and a base.

We believe that this behaviour could be the result of integrating of various technological tasks, which lead to the reconstruction of mobile toolkits. Integration of tasks was observed within modern groups of hunters-gatherers similarly extracting the landscape (Kaplan et al., 2009). Until now, this type of behaviour was mostly considered in the context of residential camps or hunting/butchering sites and workshops in quarries (Hoffecker, 2002, Shipton and Nielsen, 2015 and Vaquero et al., 2004).

The case of Pietraszyn 49a raises the question of the relationships of workshops to other places in the late Neanderthal's foraging system. It is generally accepted that shaped tools such as handaxes or knives played an essential role in butchering practices (Claud, 2014, Finkel and Barkai, 2018, Giria, 2004, Jöris, 2006, Pope and Roberts, 2005, Rots, 2009 and Veil et al., 1994). The notion that such places were functionally linked with an area of game capture and disarticulation of carcasses was also supported by microscopic studies of individual tools from Pietraszyn 49a.

Finally, in this part, we discuss issues connected to the provisioning system. At Pietraszyn 49a, discarded waste from the manufacture of shaped tools prevails over flake tools and other examples of shaped tools brought from outside (i.e. orphans; see Morrow, 1996). This scenario clearly indicates that, in the excavated part of site Pietraszyn 49, the production of transportable personal gear (i.e. formal tools) occurred (Binford, 1979 and Kuhn, 1995, pp. 22–24). However, it is difficult to assess the scale of transport. We cannot exclude that some shaped tools produced in the workshop could have been used and abandoned in the proximity. Such a scenario reflects the presence of several bifacial tools spread over the area of the Troja River valley (Fajer et al., 2001a, Kozłowski, 1964 and Wiśniewski, 2006).

To conclude, the site Pietraszyn 49a confirmed that late Neanderthals explored the northern parts of their niche using a flexible system. One of the key elements of the structure was an extraction zone covering needs connected to provisioning for food and toolkits (Fig. 15). It is clear that technological activity in the discussed place was part of a subsistence system of late Neanderthals.

Without question, the finds from Pietraszyn 49a constitute a special functional part of the land-use system. The structure of finds from Pietraszyn 49a calls to mind some camps (type A) from Crimea, indicating that a similar organisation was applied by Neanderthals over vast areas (Chabai and Uthmeier, 2006, pp. 306–307; Uthmeier and Chabai, 2010, p. 208). However, the existence of such places as workshops does not determine which mobility system prevailed at that time in front of the Moravian Gate.