Chronostratigraphy of Jerzmanowician. New data from Koziarnia Cave, Poland

Lincombian-Ranisian-Jerzmanowician (LRJ) sites are sparse, and Koziarnia Cave in Poland is one of only few such sites situated at the eastern fringe of LRJ. The aim of the recent study was to obtain new chronostratigraphic data for the LRJ industries due to their extreme scarcity in Central Europe. Although the new fieldworks did not bring new fossil directeur such as bifacial leafpoints, a detail debitage analysis enabled identifying a presence of the ventral thinning chips in layer D, which could be identified as the LRJ assemblage-containing stratum. Besides the LRJ assemblage, strata with traces of Late Middle Palaeolithic and Early Gravettian occupation were found at the site. The radiocarbon dates of Koziarnia samples show that the archaeological settlement represent one of the oldest Gravettian stays north to Carpathians. What is more, these dates demonstrate that the cave had been alternately occupied by humans and cave bears. Additionally the radiocarbon dates indicate rather young chronology of the Jerzmanowician occupation in Koziarnia Cave (c.a. 39-36 ky cal. BP). The results confirm the possibility of long chronology of the LRJ technocomplex, exceeding the Campanian Ignimbrite event.


43
Middle/Upper Palaeolithic transitional industries in Central Europe are among the most ephemeral and 44 most debated topics in Palaeolithic discourse [1][2][3]. After over 100 years of research into the subject, 45 we are still seeking for answers to crucial questions regarding, e.g. the origins [4][5][6][7][8], the chronology [9, 46 10], internal divisions [11][12][13], or even the identification of the population responsible for these 47 industries [14][15][16][17][18][19]. 48 Lincombian-Ranisian-Jerzmanowician is one of such transitional industries determined by the presence 49 of bifacially worked leafpoints made on blades obtained from double platform cores. Technological and 50 experimental studies show that one is dealing here with a predetermined technique of obtaining leaf-51 shaped blades, which were later adjusted to a minimal extent to mirror the exact willow leaf shape 52 through ventral thinning. Such technological features are present in transitional assemblages in southern 53 Poland (Nietoperzowa Cave), Southern Germany (Ranis), Belgium (Spy) and the southern part of Great 54 Britain (Beedings, Kent's Cavern), but are somewhat absent to the south of the Carpathians, where a 55 Szeletian type of industry prevails [14,[20][21][22][23]. 56 The term "Jerzmanowician" was used for the first time in 1961 by W. Chmielewski after his studies in 57 Nietoperzowa Cave located in Jerzmanowice village [20]. Chmielewski focused his research on two 58 cave sites, Koziarnia and Nietoperzowa, where the first bifacial leafpoints were found already in the 19 th 59 century. 60 In the second half of the 19 th century, the cave sediments of several sites were exploited by local 61 landlords to be sold as field fertiliser. The southern part of the Polish Jura, a karstic region rich in caves, 62 was at that time the part of the Russian Empire, but the business was driven by Prussian businessmen, 63 who organized the transit of train wagons filled with cave sediments to Prussia. In consequence, the 64 sediments of such caves as Nietoperzowa, Koziarnia and Gorenicka were heavily destroyed. Due to the 65 sediment exploitation, the original sediment level in Nietoperzowa Cave was lowered by around 1.5-2 66 m, whereas in Koziarnia Cave by around 0.5-1 m. During the cave sediment exploitation, multiple 67 prehistoric animal bones and artefacts were found. The discoveries led Ferdinand Römer, a geologist 68 and palaeontologist from Schlesische Friedrich-Wilhelms-Universität in Breslau (now University of 69 Wrocław), to study the cave sediments in detail. For this reason, he collected the already discovered 70 artefacts and conducted his own excavations at several caves in the region. The various findings 71 discovered by F. Römer [24,25] include bifacial leafpoints from Nietoperzowa and Koziarnia caves. 72 To check their stratigraphy, Chmielewski re-excavated both caves in the 1950s. In Nietoperzowa Cave,73 he found three archaeological horizons (layers 4, 5a and 6), containing in total more than 87 bifacially 74 worked leafpoints and their fragments [26]. In Koziarnia Cave, he opened ten trenches covering the area 75 of 120 m 2 in total, but none of the 21 layers determined inside the cave and on the terrace in front of the 76 cave could be clearly described as containing the Jerzmanowician assemblage. One of the layers, i.e. 77 13, which he claimed did not contain any stone artefacts, was black in colour due to the high amount of 78 charcoal [27]. Chmielewski called it a "cultural layer", and by comparing it to layers 4 and 6 in 79 Nietoperzowa Cave, he initially suggested that it was a Jerzmanowician horizon [20]. No radiometric 80 dates were obtained at that time to confirm the hypothesis. 81 Even though the determination of the Jerzmanowician culture was based mostly on the Nietoperzowa 82 Cave assemblage, the Koziarnia and Mamutowa caves were also included. A single radiocarbon date 83 (38 160 ± 1250 BP, Gro-2181) obtained for a wood charcoal from layer 6 in Nietoperzowa Cave was 84 presented by Chmielewski in 1961 [20], and since then it has been treated as the major chronological 85 framework of the whole technocomplex. It was later proposed that all the assemblages containing 86 bifacially-worked blade leafpoints from the European Plains can be merged into one category -87 Lincombian-Ranisian-Jerzmanowician (LRJ), a term widely used till today [28]. 88 The chronology of Jerzmanowician has been restudied several times since then [14,29,30,32]. The 89 analyses were in most cases conducted on the animal fossil collection. The most recent results of 90 multiple radiocarbon dating obtained on the basis of cave bear remains from Nietoperzowa Cave [33] 91 showed the limitation of the possible use of the old collection. The radiocarbon range of each stratum 92 shows all the chronological spectra observed in the cave, which might indicate problems resulting from 93 the exploration, documentation or mixing of the collection. Only a new detailed fieldwork would help 94 to resolve all the chronological issues linked to LRJ industries. 95 In order to clarify the chronostratigraphic position of Jerzmanowician, a new fieldwork project was 96 initiated in 2017. It aimed at the verification of the stratigraphy of Koziarnia Cave and obtaining reliable 97 radiometric dates for a complete profile of the site, as well as reconstructing the palaeoenvironmental 98 conditions for particular strata [34]. The paper presents the obtained chronostratigraphic data with a 99 comparison to the results published before. 100

101
Koziarnia Cave is located in Sąspów Valley, in the southern part of the Polish Jura (Fig 1). The cave 102 has a 5-metre-high entrance heading SW with the main chamber covering an area of over 100 m 2 behind 103 it and a single 40-metre-long gallery narrowing toward the end of the cave.

116
The cave was continuously in use until World War I. At the beginning of the 20 th century, when the 117 sediment exploitation was halted, a dance floor was built in the main chamber, and a resting place was 118 located on the terrace in the front of the entrance. years. All the walls disintegrated slowly causing massive damage (Fig 1.7). 139

140
The new fieldwork conducted in 2017 covered 2.85 m 2 (Fig 2). A trench was opened 40 metres from the 141 cave entrance in the NW corner of Chmielewski's trench IX in order to correlate the stratigraphy and 142 open a section in the place that could cover the highest possibly undisturbed profile. The collapsed walls 143 of the old trenches containing a mixed sediment were visible during the fieldwork. One should still take 144 into consideration that even the stratified parts of the trench might be post-depositionally moved due to 145 the slight trench wall movements. The in situ sediment was collected and wet sieved in whole with 1 146 mm sieves. A mixed sediment was collected and wet sieved with 3 mm sieves. The sieved material was 147 dried and screened in order to collect tiny microfaunal, anthracological and archaeological material. All 148 the in situ findings, including the charcoal, were 3D measured.

152
Additionally, the old collection of artefacts found in Koziarnia by F. Römer in 1879 was restudied. For 154 the purpose of chronostratigraphic studies, two unpublished bone tools were radiocarbon dated and 155 analysed through zooarchaeology by mass spectrometry (ZooMS) and traseology. 156

Radiocarbon dating
157 In each layer, charcoal was the dominated material (Table 1) Techno-typological analyses 219 The archaeological assemblage was analysed with the use of a geometric, morphometric and 220 technological approach. A set consisting of 33 features was determined for each of the pieces of 221 debitage. The attributes were divided into four general groups: 222  general artefact morphology (the size, shape, state of preservation/fragmentation, symmetry, 223 cross-section, profile, the character of the distal part), 224  the condition of the dorsal side (the direction of the scars, cortex, interscar ridges, erasing chips, 225 retouch), 226  the condition of the ventral side (the bulbs, bulbar scars), 227  the condition of the butt (the size, shape, profile, preparation). 228 In order to determine the characteristic features of the Jerzmanowician debitage, experimental studies 229 were conducted additionally. They aimed at reproducing the bifacially-worked leafpoints and studying 230 them to determine whether one can distinguish any specific features indicating leafpoint-shaping based 231 on the morphology of the debitage. During the experimental session, two blades and two flakes were 232 shaped by experimental knapper Miguel Biard into leafpoints, and the geometric morphometric features 233 of the debitage were analysed. 234

Traseology 235
Flint artefacts designated for traseological analysis were subjected to a cleaning procedure involving the 236 use of warm water and acetone. The flint material was analysed using a Nikon LV150 metallographic 237 microscope and a Keyence VH-Z100R digital microscope. The microscopic analyses were conducted 238 using with a 50x to 400x magnification ratio. The noted macroscopic and microscopic traceschipping, 239 linear wear patterns and signs of usewear, linked to changes in the surfaces caused by post-depositional 240 and utility factors. The listed traces were interpreted based on a comparison with an experimental 241 reference database, kept together with the relevant documentation at the Institute of Archaeology of the 242 University of Warsaw , as well as with reference to the appropriate literature. 243 The surfaces of the bone items selected for traseological analyses were cleaned using acetone. The 244 optical-stereoscopic Olympus SZX9 and metallographic Nicon Eclipse LV 100 microscopes were used 245 for the observation of the traces, along with the Nicon Shuttlepix digital microscope with a 6.3x to 100x 246 magnification ratio. Natural and technological traces, as well as evidence of usewear were noted on the 247 analysed material. 248

ZooMS 249
The Ambic was added to each sample. Next, the samples were incubated at 65°C for 1 h. Afterwards, 50 µl 255 of the resulting supernatant was digested with trypsin (Promega) at 37°C overnight, acidified using 1 µl 256 of 20% TFA, and cleaned with C18 ZipTips (Thermo Scientific).

267
The new fieldwork confirmed the massive destruction of the top part of the original sedimentary 268 sequence. The topmost layer in the recent cross section, which can undoubtedly be correlated with the 269 previous fieldwork, is layer K. It can be correlated with Chmielewski's layer 12. There is no possibility 270 to correlate the overlying strata J, B and A, as long as their remnants can be seen only close to the cave 271 walls. All the overlying layers have already been destroyed, mostly due to 19 th -century cave sediment 272 exploitation. 273 In general, the sedimentary sequence can be divided into four lithostratigraphic series. The lowermost 274 series is red residual clay of weathering origin (P), which has been locally preserved, a remnant of the and rounded limestone clasts (Fig 3). The upper series consists of a set of grey loams containing either 279 corroded or sharp-edged limestone clasts. It is divided into two parts by a lamina of red-brown clay 280 (layer C). Layer K', due to considerable amounts of charcoal, was a very dark black colour in 281 Chmielewski's trenches. In our trench, 40 m from the entrance, this layer still contains large charcoal 282 fragments but it is more yellowish-dark grey in colour. The uppermost part of the section, especially 283 strata situated above layer K', has been disturbed. Layers younger than layer K are preserved only as 284 remnants attached to the wall.  In total, 14 animal bones and 9 charcoal fragments from the new excavations have been dated (Table 1).

295
Radiocarbon dating was conducted on cave bear bones from the old Chmielewski excavations. 296 Additionally, the date was estimated for two bone tools and the single cave bear bone from Römer's 297 collection (Fig 4). 298

303
The results show that at least some of the strata were contaminated by recent material. Recent dates were 304 obtained solely for the charcoal fragments. This may indicate post-depositional processes connected 305 with the extended exposition of the old open trench walls to external conditions. 306 Two bones were dated with the use of U-series (Table 3). In order to check the results, one of the bone 307 specimens was dated with both the radiocarbon and U-series method. The U-series date is distant from 308 the radiocarbon one (Table 1). This suggests that U-series dating is probably unreliable at this site. The 309 reason behind this might be the open uranium system, i.e., the constant availability of uranium ions in 310 the ambient sediments, which resulted in the continuous uptake of U from the environment by bone. In 311 such cases, the U-series dates have only a terminus ante quem significance. 312 Among the newly established dates for the bones, 10 exhibit the atomic C:N ratio in extracted collagen 313 which stays within the accepted range of 2.9-3.6 [65,66] and indicates well-preserved collagen. One 314 sample (date Poz-99806) yielded too low amount of collagen to measure the C:N ratio, while another 315 (date GdA-3896) exhibited too low C:N ratio; therefore, we decided to discard these dates (Table 1). 316 The lowermost layers L-M yielded two radiocarbon dates. One of them represents recent contamination, 317 the other was established based on material coming from the 1960s excavation; thus, we are not certain 318 about its provenience. On the basis of the dating of the upper layers, layers L-M should be regarded as 319 older than ca. 47 ky cal. BP. The TL date for layer M is incompatible with radiocarbon dating. This date 320 could be biased due to the close proximity of bedrock, which resulted in a different radiation dose 321 actually absorbed by the sediment than the dose assumed in the laboratory from the measurement of the 322 concentration of radionuclides within the sample. of error for a single date from these layers. This also indicates that the erosion event followed by the re-333 deposition of layer D should be dated to around 37 ky cal. BP. 334 Archaeological assemblage 335 During new fieldwork, over 1000 stone artefacts were collected. Table 4 shows the composition of 336 artefacts found in each geological stratum. As a result of the opening of a new trench in the corner of 337 the old collapsed trench and excavating not only in situ sediments, but also partly collapsed and moved 338 layers, not all artefacts could be undoubtedly attributed to one level (  found, the other layers contained only uncharacteristic debitage, prevalently chips. To determine in 360 which layer one could expect the Jerzmanowician occupation, analysis of the chips was required. 361 The general assumption was that in the Jerzmanowician assemblage, due to the characteristic retouching 362 of the blades to shape a leafpoint, one could expect specific debitage, derived at the stage of the ventral 363 thinning. Unlike in Middle Palaeolithic and Gravettian, we assumed that in the Jerzmanowician 364 assemblage, one would be able to find chips and flakes with remnants of the ventral surface of the 365 original blank/blade. 366 To test this assumption, we conducted experimental knapping and analysed the debitage obtained during 367 the blade leafpoint shaping (Fig 5). general morphological groups. 381 The first group consists of ventral thinning chips (Fig 5E). They contain the remnants of the ventral 382 surface of the blank on their dorsal side. They come from the initial thinning of the ventral side of the 383 blade and can be described as the first generation of ventral thinning. 384 The second group contains chips which are slightly bent, with a width/length index of >1. On their dorsal 385 side, they contain scars of even smaller removals in their proximal part and a big ventral scar on their 386 distal part. Such chips come from second generation of ventral thinning (Fig 5D). Unfortunately, due to 387 the small sizes of the analysed artefacts, in many cases, it is not possible to say if the remnant of the flat 388 surface in the distal part of the artefacts is the ventral or dorsal side of the original blank. 389 The third group of chips consists of undeterminable uncharacteristic small chips (Fig 5A). They mostly 390 come from dorsal thinning and shaping, as well as secondary ventral thinning. The characteristic feature 391 in shaping a leafpoint out of the wide and rather thick blade is the presence of dorsal thinning chips, 392 which reach the interscar ridge or the blank and contain remnants of such a ridge running transversally 393 to their main axis (Fig 5C). Such flakes might be considered the specific debitage of leafpoint shaping. 394 Additionally, during ventral thinning near the butt part of the blade, in some cases a bigger chunk is 395 produced, which aims to prepare the correct angle for further removals (Fig 5B). 396 Based on experimental studies, one can assume that only the presence of chips from the first generation 397 of ventral thinning can be treated as evidence for ventral thinning, andthereforecan be associated 398 with leafpoint shaping. Although the presence of only the second generation's chips cannot be indication 399 for leafpoint production, their appearance together with first generation chips could provide additional 400 support for such an assumption. 401 Table 5 presents morphological analysis of the chips found in distinct strata in Koziarnia. The results 402 show that ventral thinning chips could only be found in layer D (Fig 6) and in the disturbed sediments 403 (n=28). In other layers, the undeterminable chips of the third type prevail. The presence of ventral 404 thinning chips leads to the assumption that layer D should be associated with an assemblage that used a 405 ventral thinning method, probably bifacial leafpoint shaping.    The upper parts of the mixed sediment provided a minimal number of Holocene period finds, consisting 416 of pottery, a glass artefact and metal objects. The ceramic assemblage is highly fragmented and poorly 417

++ -++ -
preserved. It comprises 6 pieces of uncharacteristic prehistoric pottery sherds, 1 fragment of Roman-418 period wheel-turned ware fired in reducing atmosphere, and 2 pieces of vessels dated tothe 18 th or 19 th 419 century, made of white clay and covered with yellow glaze. The find assemblage is supplemented with 420 a small fragment of a patinated glass artefact, possibly a vessel, and small pieces of undefined metal 421 objects. Due to their poor preservation, the chronology of these finds must remain uncertain.

422
Considering the recently discovered Holocene period finds, it is fair to say that they do not bring new 423 data to the studies of the use of Koziarnia in late prehistory and historical times. Some more detailed 424 insights into this topic were provided thanks to previous research campaigns, which were focused on 425 the entrance zone to the cave. As evidenced then, the site was extensively used since the Neolithic up 426 until the modern period [27]. The small amount of Holocene period finds from the 2017 excavations has 427 to be seen in the context of the distance of the trench from the cave mouth. 428

429
Both horizons are described together since layer K' has the same petrological features as layer K. The 430 only difference is the presence of a high concentration of charcoal in layer K', which changed the 431 colouration of the layer. Therefore, one can assume that layer K' is a human occupation episode within 432 the accumulation of layer K. Due to the significant destruction of the top levels, layers K and K' were 433 visible only in a tiny area of ca. 1 m 2 . The lithic assemblage consists of 139 artefacts. It should be noted 434 that materials from the layer determined ad mixed + K were also included. FLO/A45/17d) and a double microburin, reworked into a double perforator (?) (Fig 7 FLO/A45/17b). 441 Besides the tools mentioned above, one burin spall has been noted, as well as a small bladelet, which 442 could be either a burin spall or crested blade. The assemblage consists of blades and bladelets and is 443 distinct from the other assemblages due to the use of excellent quality almost translucent Jurassic flint 444 raw material. However, not much can be said about the technology and morphometric characteristics of 445 the assemblage, as mostly medial and distal flake and blade fragments were recovered. 446  the tip was observed to have a straight profile, which could be associated with hunting weapons, but 458 fracturing of this kind is not distinctive for this type of activities. 459 In turn, the second backed bladelet (FLO/A96/17) should be noted for its characteristic breakage of the 460 tip. The impact fracture (hinge terminating bending fracture) has been identified. This type of breakage 461 morphology usually enables to link the tool with hunting weaponry. In addition, the linear traces 462 observed in the middle part of the tool can be connected, due to their underdeveloped form, with its use 463 as a projectile, but simultaneously the influence of post-depositional factors cannot be excluded. 464 The third backed bladelet (FLO/A45/17a) is characterised by an impact fracture (step terminating 465 bending fracture) in its bottom part. The macroscopic morphology of the trace suggests to a certain 466 extent that the described backed bladelet might have been used as a hunting weapon. 467 It cannot be excluded that also the next specimen (FLO/A45/17c) was used during hunting. This is the 468 upper fragment of a backed bladelet broken in a unique manner; one end of the breakage with a concave 469 profile is elongated. The artefact might have been the tip of arrowhead projectile. 470

Layer D 471
Compared to other layers, this one contained a relatively rich assemblage (n=137), although the 472 materials were heavily damaged, mostly through breakage (Table 4). Among the 90 blades, bladelets 473 and flakes, only six were unbroken. Most of the flakes represent undeterminable debitage; however, at 474 least some of them have negatives attesting to bidirectional knapping (Fig 6), which is a characteristic 475 feature of Jerzmanowician [14,21]. A single flake has the features of a bifacial shaping flake. 476 On the same level as layer D, but in the disturbed sediment of the old trenches, a blade with ventral 477 thinning of the bulb was found (Fig 8: A1/17). The artefact may be interpreted as the broken part of a 478 leafpoint; however, it contains numerous post-depositional retouches, which changed its shape. The 479 usewear traces located along its longitudinal edges but due to the underdeveloped form of the polishes 480 their detailed origin cannot be determined. However, the provenience of the usewear traces is unclear.

487
To conclude, one can see at least several features indicating that one is dealing with traces of 488 Jerzmanowician occupation in layer D. The most prominent among these are the above-described 489 presence of the ventral thinning chips (Fig 6) and the debitage with bidirectional scars. A bidirectional 490 knapping scheme is confirmed also by a big blade detached from a bidirectional core found in the same 491 layer by W. Chmielewski (Fig 8: IX/17-23/61). 492

Layers E, F, G 493
These three layers represent one geological sedimentary event affected by human occupation, which can 494 be traced in layer F. This layer is characterised by a high concentration of charcoal, but only a single 495 artefact was found inside (Table 4). The above and underlying layers E and G contained in total 21 496 artefacts (Table 4). They consist of uncharacteristic elements, with a single bifacial shaping flake from 497 layer E. The cultural attribution of the assemblage is impossible. 498

499
The assemblage found in the lowermost layers consists of 436 artefacts ( Table 4). The flakes represent 500 only undeterminable debitage. The edges are heavily damaged due to post-depositional retouches 501 creating pseudo-retouched tool-like artefacts. The pseudo-retouches are present even on the 0.3-cm-long 502 chips, indicating the intensity of the post-depositional damage. At least ten chips and flakes can be 503 described as bifacial thinning debitage due to their knapping angle (60°-70°). 504 The assemblage contained one endscraper (Fig 9: A9/17), and possibly one "groszak" (Fig 9: A6/17).

505
No usewear traces were found on these artefacts. A single flake contains a multiscarred butt in the shape 506 of a chapeau de gendarme. All the described features indicate that these layers should be attributed to 507 the Middle Palaeolithic. Unfortunately, the small size of the debitage and a high post-depositional 508 damage unable more detailed cultural attributions. 509 They include 26 artefacts (Table 4) containing three retouched flakes and a single bifacial shaping chip 515 (Table 5). 516

517
Two unpublished bone tools from Koziarnia cave were found in F. Römer's collection. One of the tools 518 is a short broken piece of bone with a smoothened ending (Fig 10:1). Numerous lengthwise cracks and 519 chippings linked to exfoliation were observed on this artefact. One of the endings has been broken as a 520 result of natural factors, while the other was formed diagonally through being burnished on a stone pad. 521 No traces of usewear enabling the identification of its function were observed on the artefact. 522 The second piece is a part of a bone point with multiple incisions on its outer surface (Fig 10:2). Wide 527 linear marks of different depths, overlapping each other and parallel to the longer axis of the artefact, 528 were observed. They had been formed during the shaping of the blade through being scraped by a flint 529 tool. On the entire surface of the blade, there are distinct, deep and wide diagonal notches located parallel 530 to each other, only intersecting in the middle part of the tool. They were made with a flint flake or chip 531 through repeated sawing backwards and forwards. This type of notch should be seen as a kind of artefact 532 decoration. The surface of the artefact is smoothened and useworn, especially at its tip. 533 The spectra obtained from both bone specimens through ZooMS analysis have been taxonomically 534 identified as Elephantidae. The marker series are similar for some closely related species. In this case, 535 possible species can belong to the Elephas, Mammuthus, and Palaeoloxodon genera. Considering the 536 archaeological context, these two bone tools were most likely manufactured from woolly mammoth 537 remains. 538 Both artefacts obtained similar radiocarbon dates of 25-26 cal. ky BP (Table 1) intensive human occupation were found in the uppermost layer H' and the lowermost layer I'. 550 The second difference between the cross section presented by Chmielewski and the recent study is the 551 relatively small amount of charcoal found in layer K', which can be correlated with layer 13 by 552 Chmielewski. Nonetheless, this layer contained the highest number of charcoal fragments in the entire 553 sequence but their concentrations did not change the colouration of the stratum, as had been observed 554 by Chmielewski in trenches IX and especially IV & V. As long as these layers can be correlated with 555 human occupation, one can presume that the highest charcoal concentration could indicate an occupation 556 zone, which weakens as it nears the end of the cave corridor. 557 The comparison of all the available drawings of the cross sections from all the previously conducted 558 fieldwork enables to reconstruct the general correlation of the layers (Fig 11). Based on such correlation, 559 one can see that the trenches located in the entrance zone revealed the presence of a thick Holocene 560 sequence of humic horizons and underlying loess layers, which can be divided into two separate 561 horizons. The loess sediments, through a comparison to other caves in the region, can be correlated with   Jerzmanowicianto layers 15-16 (D-E-F-G), and the Early Gravettianto layers 13-12 (K-K'). The 580 later Gravettian episode (25-26 ky BP) cannot be attributed to any particular stratum but was manifested 581 by the presence of two bone tools found in F. Römer's collection. 582 The general correlation of the layers shows that the amount of charcoal in the Early Gravettian horizon 583 diminishes towards the end of the cave, and is the most intense approximately 20-25 m from the 584 entrance. In contrast, the thickness of the Jerzmanowician layers 15-16 increases as they near the end of 585 the cave (a 75-cm-thick layer is 40 m from the entrance), while they disappear towards the cave entrance. 586

Chronology 587
Most of the established dates cover the period between 46 and 24 ky cal. BP. However, several charcoal 588 datings provided unexpectedly recent ages (Table 1). Based on the taxonomical analysis of the charcoal 589 assemblages, it can be assumed that a part of the floated samples were indeed contaminated and this can 590 be confirmed by the presence of a few samples containing singular findings of charcoal fragments 591 belonging to fir Abies alba, hornbeam Carpinus betulus and beech Fagus sylvatica, trees that are 592 considered late-coming species in the vegetation history of Poland [69]. Such samples came mostly from 593 areas located near the previous excavations. After preliminary analysis, the existence of post-594 depositional disturbances was confirmed, indicating these places as ones that should be excluded from 595 any chronological inference. However, in a few other samples, only coniferous taxa were found (juniper 596 Juniperus communis and pine Pinus type sylvestris-mugo), which suggests that they could have 597 originated from Pleistocene layers, but their radiocarbon dating showed modern contamination (Table  598 1). This analysis has evidenced that the very meticulous study of strata in the context of all 599 archaeological and biological findings is needed to understand taphonomic processes in cave sites. 600 Another explanation for the observed discrepancy between the dates achieved from bones and at least 601 some dates achieved from charcoal, includes the altered 14 C content in charcoal fragments. From recent 602 study [70] we know that carbon in wood during the high temperature processing (such as burning) is a 603 subject to kinetic fractionation of isotopes. Namely, charcoal from coniferous wood burned in low 604 temperatures is enriched in heavy carbon and oppositely, burned in higher temperature (400-600 °C) is 605 depleted in heavy carbon in relation to the original wood. If dated charcoals became from burning in 606 relatively low temperatures, e.g. in the outer part of fireplace, this may likely produce a shift to younger 607 radiocarbon dates. 608 It is worth noting that the radiocarbon dates from layer D are not in the correct order with those from 609 the lower strata (Fig 4). This can be an effect of the mentioned isotopic fractionation in burnt wood, or 610 likely the effect of re-depositional episodes, possibly related to the erosional structures visible in layers 611 E and D. The directions of this transport are difficult to reconstruct as the 2017 excavation area was 612 quite small and delivered minimum data on the geometry of sedimentary structures. Moreover, the 613 topography of the cave floor was disturbed here due to the previous exploitation. However, the higher 614 elevation of sediments in the area closer to the entrance (especially visible in W. Chmielewski's trench 615 at the 30th metre) may suggest that this area served as a source of material for colluvial activities. If we 616 adopt the hypothesis that at least some dates from the upper layers represent a re-deposited material, we 617 need to accept that the faunal, anthracological and archaeological assemblages from these layers could 618 also have been affected by colluvial mixing. 619 If we look at the distribution of the probability density of radiocarbon dates regardless of the stratigraphy 620 (Fig 12)  that the first phase is connected with Jerzmanowician occupation, while the second and third with Early 629 Gravettian. The last human settlement phase in 26-24.5 ky cal. BP also represents traces of the 630 Gravettian occupation. Another phase, not shown in Fig 12, is the modern one (around 300 y BP until 631 modern times), based on the most recent dates achieved for the charcoal. It is worth taking note of the alternate occurrence of dates determined for charcoal fragments and animal 642 remains (Fig 12). All the dated animals were bears (mostly the cave bear, but one date has also been 643 established for the brown bear). Bears used the caves as hibernation dens, and their presence in a cave 644 could not be contemporaneous with human settlement [71,72]. Our dataset indicates that Koziarnia 645 Cave indeed had been alternately occupied by humans and bears. 646 In Fig 12,  is well-known from the younger layers 4-5-6 of Nietoperzowa Cave. It is difficult to compare the 665 sequences from both caves based on the lithology, as they represent rather different facies (Fig 13).

666
Sediments from Nietoperzowa Cave are mostly silty loams with limestone clasts, deposited in the near-667 entrance area under the strong influence of aeolian activity. In Koziarnia Cave, the recognized sediments 668 are coarser, they were deposited quite deep inside the cave, and they are mostly limestone debris. 669 However, the proportion of angular to subangular clasts, presented by Madeyska from southern Poland (Fig 14) . The chronology of the Jerzmanowician assemblages is currently based 685 on 38 radiocarbon dates [S3, 33]. The majority of the radiocarbon datings were obtained on either cave 686 bear (n=29) or bird (n=4) bones without human activity traces. Taking into consideration the fact that 687 cave bears and possibly also birds of prey did not cohabit the caves with humans, in order to determine 688 the chronology of human occupation, we shall instead rely on radiocarbon dates provided by charcoal, 689 bones with cut marks, or animal species which do not naturally live in caves, such as mammoths. The bone points from Mamutowa Cave are older than the oldest available Aurignacian dating north of 723 the Carpathians, but at the same time they overlap with the Jerzmanowician settlement in Koziarnia and 724 Nietoperzowa Caves. One can, therefore, assume that they could have originally belonged to the 725 Jerzmanowician assemblage in Mamutowa Cave. Unfortunately, the chronology of Mamutowa Cave is 726 mostly based on radiocarbon dates made on cave bear and bird bones. The sets of dates from underlying 727 and overlying strata as well as from layer VI indicate some post-depositional sediment mixing (S2). 728 The possibility of the long chronology of the LRJ technocomplex exceeding the Campanian Ignimbrite 729 (CI) eruption event is confirmed also by results obtained in Lincombian sites e.g. Beedings, indicating 730 its lasting up to even 30 ky BP [108]. 731

732
It is worth noting that aside from the Middle Palaeolithic and Jerzmanowician, also two Gravettian 733 occupation phases can be identified in Koziarnia Cave. The first one could be correlated with the second 734 and third human settlement phase recorded by the charcoal dated to 35-31 ky cal. BP (Fig 14). This 735 dating overlaps with the earliest traces of the penetration of Gravettian hunters, confirmed recently in 736 Henryków 15 in the Sudetes piedmonts. The chronology of the settlement traces in Koziarnia Cave 737 indicates that the earliest Gravettian groups penetrated not only the nearest vicinities of the Moravian 738 Gate but went much further into the Polish Highlands. 739 The second settlement phase, which is recorded only by two bone tools of uncertain stratigraphic phases in the cave. We may, therefore, assume that the cave was occupied in the Late Middle 750 Palaeolithic, but the typo-technological character of the assemblage is still to be discussed. The site was 751 also occupied during the Middle/Upper Palaeolithic transition. This occupation phase can be identified 752 as Jerzmanowician and should be dated to 39-36 ky cal. BP. The obtained radiocarbon dates indicate 753 that the Jerzmanowician tradition lasted longer and did not finish with the Campanian Ignimbrite 754 eruption. Above the Jerzmanowician strata, a thin sterile layer can be observed, separating the overlying 755 Gravettian strata. The earliest Gravettian occupation can be dated to 35-31 ky cal. BP, and thus 756 represents the earliest Gravettian occupation in the Polish Jura, and one of the earliest to the North of 757 Carpathians. 758 One should also emphasize that the recent results confirm the previous assumptions, claiming that 759 humans and animals did not cohabit caves, even if their traces are found in the same lithostratigraphic 760 layers. For this reason, only radiocarbon dates obtained either on charcoal fragments or modified bones 761 or teeth should be used for determining human settlement at cave sites.