Osseous tools and personal ornaments from the Epigravettian sequence at Badanj


 The Late Upper Palaeolithic (Epigravettian) sequence at Badanj has yielded an important dataset about the occupation of the hinterland of the Eastern Adriatic catchment zone in the late Pleniglacial. The site also harbors one of the rare occurrences of Upper Palaeolithic parietal “art” in southeastern Europe in the form of a large rock engraving. Another notable aspect of the site is the presence of engravings on portable objects made from bone. The first excavations at Badanj, conducted in 1976–1979 in the zone around the engraved rock, yielded a surprisingly large number of personal ornaments (over 1000 specimens) from a variety of primarily marine gastropods, scaphopods, and bivalves, and red deer canines. Here we review what is currently known about the site and report our preliminary findings from the study of the collection of personal ornaments as well as osseous tools, some of which were marked by regular incisions forming decorative motifs. We also report two new direct accelerator mass spectrometry (AMS) dates on antler barbed points.


Introduction
Studying the Late Upper Palaeolithic period along the rim of the Adriatic Basin offers one of the best opportunities to examine resilience and change among prehistoric foragers over time in their responses to some of the rapid shifts in climatic conditions towards the end of the Pleistocene. These climatic shifts and ensuing habitat losses signi cantly altered landscapes and directly affected mobility patterns of possible depiction of the rear right side of a large quadruped interpreted by Basler as a horse with its right hind leg shown straight, the thigh and sti e shown in the foreground and also possibly the left hind leg (none of the hoofs shown) less deeply engraved and left oating as if depicting a movement of the animal. Basler 10 maintains that before the upper levels with recent goat dung and the top archaeological layer were removed, the boulder was covered by archaeological levels from the later Epigravettian use of the shelter. He also suggested that people might have lit re over the engrave boulder in antiquity. The leg and back parts of the outline were stressed by deeply carved lines (up to 5 mm) while the area inside the outline was further lled with multiple longitudinally engraved lines some of which also seem deeply carved. There is a possible representation of male sex as well as "arrowheads" that might have pierced the stomach of the depicted animal. Basler 11 speculated that the rock might have served as a place of initiation of young hunters.
The rst archaeological excavations at the site were conducted by Đuro Basler, a curator at the National Museum of Bosnia andHerzegovina in Sarajevo, from 1976 to 1979, covering around 50 m 2 of the eastern part of the shelter. Apart from one article written in 1976 about the discovery of Badanj, with descriptions of the engraved boulder and its signi cance, 12 and a later, relatively short, summary of int industry, osseous tools, and ornaments provided by Đuro Basler in his coverage of the Palaeolithic of Bosnia and Herzegovina in the synthetic volume Praistorija jugoslavenskih zemalja I: paleolitsko i mezolitsko doba, 13 there is no detailed published information about the rst investigations of this site. Furthermore, the National Museum of Bosnia and Herzegovina in Sarajevo does not possess detailed eld records of the investigations made by Basler at Badanj, apart from the collection of inventoried material collected at the site and the inventory books with information about the provenance of inventoried nds. Despite the current uncertain state of eld documentation from Badanj, accounts of those who participated at the excavations in 1976-1979 con rmed that the sediment excavated at the site was transported for dry or water sieving (Fig. 5), 14 which might have contributed to the large number of recovered int artifacts as well as beads (see below). Arbitrary cuttings were 10 cm thick based on depth measurements of each artifact provided in the museum inventory book.
In 1986-1987, the second phase of research at Badanj commenced in the framework of a collaborative international project entitled "Palaeolithic-Mesolithic Occupation of the Adriatic-Mediterranean Zone of Bosnia and Herzegovina," led by Zilka Kujundžić of the National Museum of Bosnia and Herzegovina in Sarajevo and Robert Whallon of the Museum of Anthropology and Department of Anthropology at the University of Michigan, Ann Arbor, USA. 15 The area investigated during the 1986-1987 excavations encompassed the western part of the site, exposing a surface of around 35 m 2 (Fig. 6). The investigated 10-m-long and 3-4-m-wide transect ran from the back wall of the shelter to the front edge of the site, beyond the dripline (the limit of the area of the site sheltered from elements by the overhanging roof) (Fig.   7). Excavated layers were primarily distinguished by following sediment and texture changes in the sediment matrix and rarely by arbitrary cuttings. The smallest spatial collection unit was one-quarter square meter (0.5 by 0.5 m) within a 1 by 1 m grid. All sediments were water-sieved using 3-mm wire mesh-screens with a subsample containing several liters of excavated sediment selected for otation and ne water-sieving with nylon stockings. 16 Stratigraphy There is not much information currently available on the stratigraphic sequence unearthed during Basler's excavations at the site in 1976-1979 apart from information that the depth reached in various parts of the trench he excavated around the engraved rock was from 1.5m to 2.3 m deep, measuring from the surface encountered in 1976. 17 Relative depths of artifacts could be used as very coarse-grained indicators about early versus later parts of the stratigraphic sequence in making very broad comparisons with stratigraphic units and their characteristics more securely established during the second research phase (see below). The reached depth of up to ca. 2 m of archaeological deposits, judging by the provenance of artifacts found at this depth, suggests a deep stratigraphic sequence in this central part of the shelter.
In the course of the second research phase in 1986-1987, some 29 archaeological layers were recognized in a meter-deep trench, with the depth often varying from 80-90 cm only. 18 The thickness of layers varied from only few centimeters to several thicker layers of 10-15 cm. A representative published stratigraphic section is provided here (Fig. 8) as well as the available description of the physical properties of each of the discovered units along with the available information about features and material culture recognized within each of the units (Table 1). There is a gentle dip of all layers towards the back wall of the shelter, from just outside the present-day dripline, probably following the slope of a natural depositional cone formed by eroded rubble, i.e., sterile éboulis from the shelter roof. There was a large hearth complex, found towards the back of the shelter, which both cuts through and inter ngers with various levels. There are at least four major divisions within this continuous hearth sequence. It has been suggested that there is a considerable degree of vertical mixing in the area of the hearth, and it has been excluded from the published analysis of lithics 19 and faunal remains 20 from the site. The continuous reuse of the hearth basin suggests it was likely regularly re-cut, raked out, and re-deposited with mixed sediments, making the links between the hearth and occupation levels hard to establish with certainty. Apart from the hearth complex, the excavators have divided the recognized levels into two groups. Levels of relatively limited spatial extent (<4 m 2 ) were con ned to the back of the shelter. The rear layers close to the shelter wall were generally looser and often with more abundant artifacts than the layers closer to the front of the shelter. This is particularly true of the lower portion of the sequence in the back of the shelter ( Levels 7,7a,11,12,16,and 17). Some of the layers were likely, at least in part, waterdeposited (Levels 5 and 5a). The rear layers are seen more as refuse areas while the front layers, basins, and pits can probably be best interpreted as individual episodes of occupation. The front levels were more extensive horizontally (around 10-15 m 2 ) and thin out towards the south, likely representing occupational oors. The provenance of discovered nds is marked rst by the designation of the 1 by 1 m quadrant as a combination of a letter and a number referring to the grid shown in Fig. 6, followed by a letter of the subquadrant (A, B, C, or D), and nally a number or combination of a number and a letter designating the level. 21 Table 1. Description of excavated levels at Badanj, 1986Badanj, -1987 excavations. Modified after Whallon 1988 cited by Miracle 1995, Appendix I. Numbers in square brackets in the fourth column indicate the number of identified flint tools, cores and core by-products found in each of the unmixed levels (after Whallon 1999, Table 31.1). Level Extent, color, and consistency Features Archaeological material 1 Generally gray in color, varying to gray-brown or brown in a few places, and showing only occasional spots of a more reddish cast. In most areas, the gray or gray-brown earth of this level was mixed with a moderate amount of small, angular rocks (éboulis).
A moderate amount of archaeological material, representing a mixture of several periods, resulting from disturbance of this deposit, partly by what appeared to be modern ash and fireplaces.
[431] 2 Clearly distinguishable as a relatively soft, loose, red, reddish, or reddishgray deposit, containing less angular éboulis than Level 1. Near the rear, this level was more fine-grained, humic, and more moist, changing in color to a brown or yellow-brown (sq. I-K, 5-6), although this occurred as a gradual gradation from the characteristic red color of the level further to the front, rather than as an abrupt change. At the very front of the site, the typical red of this level became a reddish gray and finally gray.
In a number of areas where ash was found to be mixed into the earth of Level 2, there were found occasional patches of ash and ash mixed with snail shells. However, these appear to originate largely in earlier levels and to have been incorporated into Level 2 by disturbance of these earlier deposits during the occupation of Level 2. The top of one of the major features of several earlier occupations (perhaps all in this area of the site), a large, deep, hearth complex, appears already at the base of Level 2 (sq. J-K, 6-7).
Moderate amount of archaeological material, but different from Level 1, it appears to be in situ for the most part. [621] 5 Limited to the northeast corner of the excavation area, near the back wall of the shelter (sq. J-K, 5-6). Fine, yellowbrown silts, with no, or extremely small number of éboulis inclusions. It looks very much like a water-deposited sediment.
Level 5 clearly overlay not only the large hearth complex but also Level 2a. [155] 2a Consists almost entirely of a large concentration of soft, gray ash. In a few cases, the deposit seems to consist primarily of soft, gray ash, with few or no snail shells, but the inclusion of moderate quantities to abundant masses of snail shells, many of which are unbroken, is typical of this layer. In one or two spots, particularly in sq. I5, against the back wall of the shelter, there is more earth in the deposit, and it is a brown to dark brown color, although it remains soft and full of snail shells.
The bulk of this concentration occurs in a large, relatively deep pit, or complex of pits, to the rear of the shelter (from row 8 to the rear) and particularly in the northwest, rear corner of the excavation area (sq. I5-7, J5-6). The very last phase of intense use of the large hearth complex in sq. J-K, 6-7 appears probably to have been associated with Level 2a. 2b Level 2 deposits were so deep in sq. I5-6 that a lower division was created. This division was at first entirely arbitrary, but in looking at the remaining rear wall profile along the 4/5 m line, these lower ash and snail shell deposits look reddish in contrast to the real gray, which is characteristic of Level 2a. This difference may not be greatly significant in depositional/occupational terms, and it otherwise looks as though 2a and 2b together represent a single occupation phase.
Snail shells, flint, and bone. [104] Hearth Stratigraphic divisions within the hearth complex were distinguished primarily on the physical character of its deposits.
In the course of the building up of this complex, there appear to have been stages during which particularly long or intense fires were made that created heavily burned and fused deposits. Some four such phases were observed, which serve to divide the complex into four major divisions: The Top Hearth, Upper Hearth, Lower Hearth, and Basal Hearth.
The nature of the materials within the hearth does not appear to vary significantly throughout the sequence of its use and re-use. 3 A very red layer -sometimes being a bright red, but more usually occurring as simply a red or reddish, light, soft, fine, even powdery deposit, with relatively much éboulis. Level 3 grades at its outer edge from red to gray, brownish gray, or gray-black, usually slightly darker (or a bit more compact) than Level 2.
To the rear, Level 3 has been cut into by the pits of Level 2a. [587] 5a Reddish brown to brown, silt deposit in the rear of the shelter. It was generally distinguishable from the overlying Level 5 by its darker color, although in a number of places this distinction was very difficult to make in excavation, leading to some possible confusion of the two in these spots. The nature of the sediments in this layer varies vertically. At the top, Level 5a is uniformly pure silt, with little or no éboulis or snail shell inclusions. However, irregularly, here and there, it was found to become mixed with some éboulis, snail shells, etc., being more compact, also, in these parts. Elsewhere, both above and below such areas, it remains pure silt. At the very bottom of the level, over most of its area in sq. I-K5, it was filled with large masses of very light, porous but compact, yellow (and occasionally gray) material.
Level 5a clearly interfingered with deposit of the hearth. An interdigitation of Level 2a deposits with the upper part of Level 5a can be seen in the 4/5 m profile at the rear of the site. Level 5a becomes extremely thick in places, in conformity with the interpretation that it represents a long sequence of accumulation continuing during the deposition of more than one other stratigraphic unit beside (west) and in front of it. This is tentatively identifiable as redeposited loess, washed in from the surface on the hill above the shelter, and evidently represents some rather specific depositional/climatic conditions near the beginning of the accumulation of this layer. Generally speaking, though, the predominantly silt composition of Level 5a seems to argue, as in the case of Level 5, for water as the agent of deposition here and an interpretation of this layer as a natural (geological) accumulation of material washed into the rear of the site, rather than an occupation level (or midden deposit). [170] 4 Mostly reddish-brown in color, with frequent variations toward reddish gray on the one hand, or, less often, Where Level 4 meets the hearth complex, it does not interfinger with hearth deposits, which seem, rather, to cut into the level Lots of flint and bone.
[486] darker red. The front edge of Level 4 became darker, being brownish gray to dark gray, for some 0.5-1 m prior to its termination against sterile, yellow éboulis. The sediments of Level 4 were typically unconsolidated and contained relatively large amount of éboulis, and occasionally snail shells, especially in sq. I6-7.
forming a basin (as also is the case with Level 3).

7
Confined to the northwest rear corner of the excavation area. It was a very loose, dark, gray to almost black deposit, with perhaps fewer éboulis inclusions than Level 4 or 5, but of a less silty, coarser texture than Level 5a. It was clear in excavation that Level 7 terminated tapering out on top of underlying Level 6, seemingly ending originally thus, rather than having been cut into and truncated or removed by later occupations.
In character it looked very like an occupation layer, similar to the more extensive ones that extended fully forward to the sterile éboulis at the front of the site. Restricted to the northwestern corner of the excavation area, underneath Level 6. It was a very loose, unconsolidated deposit, very dark in color, being very dark gray, gray-black, to black, and being described at points as "greasy." Given its size and location, it is impossible to think of this level as a true occupation layer in the same sense as the horizontally more extensive levels. brown color. It is also a generally thicker layer. cut into cleanly by the hearth complex as is Level 8. 10 Definitely darker and more gray in color. It was similarly loose in texture, containing a light to moderate amount of éboulis. It was of variable thickness, and was extremely thin in places. Level 10 covered about the same area as overlying Levels 8 and 9, but it was unusual in reversing the trend of the steady retreat of the front edge of lower stratigraphic levels. It was noted in excavating this layer that in some places a noticeable amount of light brown sediment like that of Level 14 was mixed in patchily with Level 10.
This level exhibited a variable relation to the Basal Hearth, interfingering with it in some places, but apparently cut through by the hearth elsewhere.
Occasionally, relatively large amounts of flint and bone. [435] 14 Predominantly yellow-brown to light brown in color, clearly distinguishable in this respect from the overlying Level 10. Moderately compact, rather than loose and unconsolidated like the overlying level, although it did contain moderate to abundant amounts of small éboulis. In extent, this level was restricted to the front of the excavation area.
Along much of its rear edge, it was cut into and truncated by the Basal Hearth. However, along the western side of the excavation it could be seen that Level 14 thinned out and ended naturally about half-way back into the shelter. In the front, this level began once again the progressive retreat of the outer edge of the site. [248] 11 Found in at the rear of the shelter, particularly in the northwest corner of the excavation. It was characteristically a red to dark red layer, although it darkened to a redblack or even black color in a few small areas. The sedimentation in this layer ranged from loose to moderately compact (always more compact than Level 7a above it) and were somewhat silty.
It appeared clearly to be a cultural deposit, rather than of geological origin, although not a major occupation area, given its restricted size.
Level 11 consistently contained a notably high density of bones and flints. [434]

11M
Has the appearance of a highly mixed deposit (thus the designation "M"), consisting of a loose, silty, gray to graybrown, ashy deposit. Level 11M was found in back of Level 11, against the shelter wall, and to the east of Level 11, in the northeast corner of the excavation, behind the hearth complex.
The transition between levels 11 and 11M was rapid, but not sharp. There seemed to be always some intermingling of the two levels, which led to their designation by the same level number, strongly differentiated, however, in terms of the nature of their sediments and what they likely represent depositionally. excavation, reaching at its surface as far, or almost as far, forward as Level 14 above it, but showing at its base a significant retreat of the area covered by archaeological layers over this part of the site. It was a relatively thick layer, of brown earth, which was darker toward the rear of the shelter, stained slightly red near its contact with Levels 16 and 17, becoming lighter and more gray-brown toward the front. There was a great deal of small éboulis included in the sediments of this layer, and on the east side of the excavation, these éboulis were strikingly white in color, rounded, and soft, giving a strong impression that they had been heavily weathered.
variable, being intact underneath the hearth in places, but clearly having been cut onto here and there by the hearth as well. At the rear of the shelter, Level 13 abutted Levels 16 and 17 directly, all three levels appearing to be stratigraphically equivalent, but with rather sharp boundaries among them all, with no evident intermingling or interfingering of their respective deposits. Level 13 seems to be a broad occupation layer, and perhaps Levels 16 and 17 can be seen also to belong to a complex of levels that can be considered together to represent the deposits left from a single, major phase of occupation of the site.

16
Restricted to the northwest corner of the excavation area. It was dark brown to brown-black, with a reddish cast to it. It was very unconsolidated and filled with a great deal of small and larger éboulis.
It matched Levels 13 and 17 stratigraphically, but abutted them, as described above, rather than intergrading with them.
Abundant flint artifacts and bones, which appeared during excavation to be larger in this layer than usual.
[193] 17 A relatively small unit in the northwest, rear part of the excavation, occurring in back of the hearth and between it and Level 16 to the west. The sediment of this layer was distinctively red, fine, and very soft, often powdery. It contained little éboulis.
It contrasted sharply in various characteristics with Levels 16 and 13, although it appeared to lie stratigraphically adjacent to these levels rather than to over-or underlie either of them.
A good deal of smaller archaeological material, and, near the rear wall in sq. K5, many whole snail shells. [125]

Absolute Chronology
There are currently only ve available radiocarbon dates from Badanj (Table 2). Two measurements were obtained immediately after the 1986-1987 excavations and were made on burnt bones. OxA-2196, from bottom-of-the-sequence Level 13, possibly provides a representative date for the start of the sequence in the Oldest Dryas ( Fig. 9). OxA-2197 comes from the middle part of the sequence, Level 6, and falls into the duration of the Bølling interstadial. 22 In the mid-1990s, another burnt bone sample from Badanj was submitted to Oxford for dating by Geoff Bailey, along with samples from Palaeolithic sites in Epirus. 23 The sample came from Level 4, which is younger than Level 6 in the stratigraphic sequence, but the obtained result (OxA-5895, Table 2) was surprisingly early, corresponding with the date earlier obtained for Level 13. The only explanation of this discrepancy is that some residual material from earlier levels ended up in the assemblage of bones in Level 4. Future, more robust dating and careful selection of dating material (e.g., articulates or matching unfused epiphysis and diaphysis of young animals, both indicative of freshly deposited carcasses that minimally moved after deposition) from various levels at the site will allow for the building of a more reliable absolute chronological framework.
As part of our analysis of the osseous material from Badanj (see below), we have submitted nine samples for dating to the Oxford Radiocarbon Accelerator Unit in September 2018 and August 2019. Apart from one sampled barbed point (inv. 10016, Table 1) and two unmodi ed bones from levels 2a and 3, which came from the 1986-1987 excavations at Badanj, the rest of the samples come from the collection unearthed by Basler in 1976Basler in -1979. Two samples were taken on two incised bones (Fig. 11) and both were withdrawn due to the low collagen yield (%N 0.22 and 0.25 respectively, i.e., below the Oxford Laboratory's minimum threshold). One sample (P-48056) was taken on a perforated red deer canine ornament ( Fig. 17:10) and its dating failed, while one sample on an unmodi ed red deer molar from the depth of 1.3-1.4 m in quad. XXII/II produced OxA-39631, which provides the range of 14,310-13,310 cal BP (95% con dence). Another two samples were taken from two specimens of barbed points (possible harpoons but see below for more details) (Fig. 12), and both produced comparable dates (Table   2). After the calibration using OxCal v.4.3.2, 24 OxA-38111 gave the range of 13,310-13,110 cal BP (95% con dence) and OxA-X-2796-45 gave the range of 14,220-13,310 cal BP (95% con dence), with the two measurements not overlapping. The precision of OxA-X-2796-45, which has large error terms, was somewhat affected due to the low pretreatment yield of collagen (%N=0.43, i.e., below the Oxford Laboratory's minimum threshold), even though it is in all likelihood earlier than OxA-38111. While the barbed point dated by OxA-38111 has no stratigraphic information, it is very likely that similar to other barbed points discovered during the rst and second excavation phases, it also comes from upper levels in the sequence of the site (see below), and thus represents a date of the nal Epigravettian use of Badanj. The barbed point inv. 10016 dated by OxA-X-2796-45 comes from Level 6 in quadrant J5B next to the hearth complex in the back of the shelter, and is in a broad agreement with another previously obtained measurement (OxA-2197), which came from the same level. One lower left human canine found in the Badanj collection with no contextual information was also directly AMS-dated, producing a Late Roman period date -OxA-39697 provides the range of AD 255-415 (Table 2).
The existing and newly available dates might be representative of the span of the Late Upper Palaeolithic use of this site. However, there is an increasing need to provide a more robust chronological framework for the occupation of Badanj, and this can be achieved by planned future targeted AMS dating of preserved organic remains from the site.  There were in total 7860 knapped stone artifacts from the 1986-1987 excavation seasons, excluding the material found in likely mixed contexts of the hearth complex and Level 11M. Analysis of the lithic material from the 1986-1987 excavations has shown that within the general "Epigravettian" lithic industry ( Fig. 10), widespread across Mediterranean Europe from southern France to Greece, there are several diachronic trends, suggesting at least two distinct phases in the sequence -early and late, with the transition occurring roughly in the middle of the sequence between Levels 6 and 7a. 27 The following observations have been made about marked differences between these two periods: (1) predominance of backed bladelets with straight or slightly curved backs in the Early period in contrast to a predominance of thumbnail or "circular" scrapers in the Late period; (2) backed bladelets with noticeably curved backs are more a characteristic of the Late period; (3) backed blades and both small and large backed ake points are more frequent in the Late period; (4) geometric microliths (mostly crescents/segments, with about half as many triangles) are exclusively restricted to the Late period; (5) there is a decline of end-ofblade scrapers and large, massive sidescrapers from the Early to Late period; (6) truncations are more common in the Late period. 28 Apart from these trends, one could also note that there is only one reported shouldered piece for the whole assemblage excavated in 1986-1987. 29 Well-prepared prismatic or pyramidal cores are rare, often these are poorly prepared and highly exhausted. There are also abundant splintered pieces (pièces esquillées), made by knapping on hard surfaces with bipolar aking techniques, which are interpreted as a form of cores for microblades. A spatial analysis of the distribution of lithic nds for representative Level 6 suggests differences between the rear and front of the shelter: a range of tools, such as thumbnail endcrapers and borers in the rear of the rockshelter, close to the area of the hearth complex in the well-sheltered zone, suggests a range of domestic, small-scale craft activities, while end-of-ake scrapers and likely armatures (microgravettes, segments) were concentrated in the front of the shelter, possibly suggesting hide and skin processing activities, as well as activities related to the repair of hunting gear. 30 Comparison of the Badanj lithic assemblage with those in adjacent regions provides close parallels between Badanj and the assemblages in Apulia, southern Italy. For instance, the same trend seen at Badanj with regard to the dominance of backed bladelets in the early Epigravettian phases in contrast to the dominance of thumbnail or "circular" scrapers in the later phases is also found in southern Apulian sites (e.g., Romanelli, Taurisano, Fig. 1). The other similarity between the two regions is the recurrent presence of splintered pieces in both regions. One of the key differences between these assemblages, however, is the absence of microburins at Badanj and their persistence in the Italian Epigravettian assemblages. It is interesting to note that similarities in the character of the lithic assemblage between Badanj and southern Apulia are not shared with northern Apulian sites, such as Grotta Paglicci, in the Gargano peninsula (Fig. 1 The general trend over time from earlier to later levels is the increase in the frequencies of wild boar and roe deer, and a drop in the frequencies of chamois/ibex (from Level 6 onwards). The red deer is steadily present as the staple hunted game species from the very start of the sequence but its dominance increases further from Level 6 onwards. There is a sharp rise in the hunting of roe deer at the end of the sequence (Levels 2 and 1). These major shifts are explained by environmental changes due to the expansion of forest cover at the expense of angry karst, the latter being a preferred habitat of chamois/ibex. 47 This would be expected from warmer climatic conditions of the Bølling/Allerød interstadial into which the later part of the Badanj sequence can be placed (Fig. 9). A small number of identi ed marmot (Marmota marmota) remains are restricted to the lower part of the sequence, which would correspond to the assumed duration of the Oldest Dryas. On the other hand, throughout the sequence, red deer, roe deer, and wild boar remains were more frequently found in the back levels of the shelter while chamois/ibex remains were more frequently found in the front levels. Based on his analysis, Miracle 48 suggested grouping of different levels in the sequence of Badanj into Upper (Levels 1, 2, 2a/2b, 5), Middle (Levels 3, 4, 5a, 6, 7), Lower (Levels 7a, 8,9,10,11,12,14), and Lowest (Levels 13, 16, 17).
In some levels (2a, 5a, 4, 12, and 17), large concentrations of land snails (Helix sp.) were found, often next to the back of the shelter. By far the largest concentration was found in Level 2a, and it remains unclear to what extent the accumulation of land snails relates to human or environmental factors. 49 Very few sh specimens were recovered at Badanj (NISP=21) and the discovered remains are primarily concentrated throughout the Later period of the site sequence, which would correspond well with the appearance of barbed points made from antler, which might have possibly been used as harpoons for shing in the nearby river around the same time (but see below). Yet, it is unclear whether such a small assemblage of sh bones would justify the presence of barbed points, which also could have been used in targeting game, and we will return later to a discussion of this category of tools and their possible function. Bird bones are similarly rare (NISP=16).
Miracle and O'Brien 50 examined two types of seasonality indicators in order to suggest the season of occupation of Badanj against the background of existing models of Palaeolithic foragers' seasonal transhumance and likely changes in patterns of forager logistic mobility in periods of signi cant climatic unpredictability towards the end of the Pleistocene. There were signi cant and sometimes rapid climatic and environmental shifts during this period that might have had a knock-on effect on the seasonal density and availability of animal herds and displacements of human population from the presumably productive coastal plain areas, caused by sea level rise, into the hinterland of the Adriatic Basin. Fetal remains of red deer and tooth cementum annulation (TCA) on red deer teeth were examined, considering that red deer is the most dominantly present and thus representative subsistence staple species. While sample sizes for neither of the two methods are particularly large when specimens are disaggregated by different levels of the sequence that spanned several millennia, at face value the data seem to suggest a strong pattern of winter use of the site for targeting hind (female) herds, especially in relation to the Early period (Levels 7, 13) of the sequence. There seems to be a shift in the Late period (Levels 2b, 3, 5a) of the sequence with more variable seasons of occupation as mortality pro les for red deer shifted into the summer and/or fall, which could be in line with changes brought by the onset of the Bølling/Allerød interstadial around 14.6 kya cal BP, the rise of sea levels, and the assumed general shifts in the scheduling of the seasonal rounds of the wider regional resource exploitation and mobility. The targeting of female herds by Badanj hunters in the winter might have stemmed from the decision to obtain prey of relatively fat-rich individuals compared to malnourished male individuals of red deer during the same season. 51 The relative rarity of red deer antlers, restricted to male individuals, throughout the sequence of the site may corroborate the scenario of the preferential hunting of female herds. On the other hand, the presence of some tools made from red deer antlers from Level 6 onwards (see below) suggests that red deer antlers used in manufacturing these artifacts might have been collected as shed antlers, that some male individuals were hunted after all, and/or that these barbed points might have been made elsewhere and brought to the site as nished products.
We will return to a further discussion of the relevance of the examined indicators of the seasonality of site occupation when modeling and reconstructing a wider pattern of forager residential/seasonal mobility in this part of the Adriatic Basin. This discussion will particularly be relevant when considering the availability, use, and frequencies of certain marine gastropods, scaphopods, and bivalves used as personal adornment in the rich assemblage of these items from Badanj.

Materials And Methods
We have examined the largest part of the existing collection of osseous tools and personal ornaments from Badanj but our analysis is not completed at present. Hence the data we present below are preliminary insights with much more detail and a comprehensive coverage to be provided elsewhere. All examined specimens were measured, photographed in several projections with a NIKON D3200 digital camera equipped with macro lens, studied by naked-eye, and further examined using a Colestron Microdirect 1080 HD handheld digital microscope with magni cation up to 220x in order to understand feasibility for further techno-functional study. While use-wear and residue analyses are somewhat hampered by the heavy presence of consolidants over the surfaces and writing of museum inventory numbers, a number of artifacts have already been selected for further microscopic examination using both low and high magni cation. In addition, in the past, for exhibition purposes, a large number of personal ornaments were strung together, which, for the moment, in some cases prevents their detailed examination. While the osseous materials we examined include specimens found during both 1976-excavations only.

Osseous Tools
Basler 52 reports a total of 11 osseous tools from the zone he excavated in the late 1970s. In the museum collection, we have found at least 53 osseous tools that come from Basler's excavations at the site. In the course of the 1986-1987 excavations, osseous tools were also found in low frequency compared to int implements. It has been noted that stratigraphically the rare occurrence of antler barbed points is restricted to the Late period in the sequence, i.e., from Level 6 onwards. 53 Pointed tools A number of bone tools in the examined assemblage are distal parts of pointed tools (Fig. 11). As with other artifacts, doing use-wear analysis is to some extent hampered by the use of consolidants. Based on manufacturing traces, slightly asymmetrical pointed tools, made on medium to large mammal long bone/metapodial diaphysis, are created by invasive int scraping of the surfaces. No earlier steps in the production sequence can be observed on the specimens we have examined. A large asymmetrical point with a broken tip was likely made from the ulna of a large mammal (Fig. 11:10). Asymmetric pointed tools made from bone are known from our examination of the assemblage of osseous tools found in the Early to Late Epigravettian-dated Layers VIII and VII at the Crvena Stijena rock-shelter in Montenegro (unpublished data). Similar basic shapes can also be found in the assemblage of Early (Layer C/d) and Late (Layer B/d and Complex B) Epigravettian tools from the site of Šandalja II in Istria, Croatia. 54 Further analogies can be made with the assemblage of bone tools from the Dalmeri rock-shelter in northern Italy (Asiago Plateau), where, along with asymmetric points, numerous projectile points were found as well as blanks made from red deer antler. 55 There is one tapered point made from a long bone diaphysis ( Fig.  11:7) and one symmetric bipoint made from an antler blank ( Fig. 11:8). Both of these specimens show A large and fully shaped antler blank (Fig. 11:9), manufactured by double longitudinal grooving along the axis and transversal cutting of its lower, proximal end, is noteworthy. This is likely a roughout, possibly intended for the production of a long projectile point or barbed point. The stratigraphic position of the specimen, which was found in the course of the rst excavations at the site at the depth of 1.0-1.1 m, although relatively deep, could still likely be considered roughly at the boundary between the Early and Late Period of occupation of the site as identi ed in the area excavated in 1986-1987. Hence it could relate to the production of barbed points made from antler.

Barbed points
All of the examined unilateral barbed points (Fig. 12) were made from red deer antler. There were two to three or more barbs on each specimen. One heavily fragmented specimen (Fig. 12:5) does not have preserved barbs, but we include it in this category on the basis of several traits-it was made from red deer antler, oblique transversal cuts on one side and speci c use-fracture indicate that a barb might have been present here, and it has a perforation near its distal end-all of which may tentatively suggest it was a barbed point. The specimens exhibit different morphologies. They are well preserved even though they are affected by taphonomical alterations-discoloration and exfoliation of the surface tissue caused by water action, as well as burning, the latter clearly visible on two specimens ( Fig. 12:4-5). No manufacturing traces related to the debitage can be discerned on these specimens as these were scraped out while shaping them, with the exception of one specimen on which longitudinal grooving marks in the shape of deep striations are still visible on the side of the tool where barbs are missing ( Fig. 12:3).
Overall, double grooving was used for carving all the blanks, which were later regularized by scraping. Barbs were added at the end of the manufacturing process by tracing them rst and subsequently cutting them using a int tool . A large perforation is visible on two artifacts-on one entire specimen where it was carved and, subsequently enlarged by rotation ( Fig. 12:1), and on the partly burnt specimen ( Fig. 12:5) where it was carved by deep grooving. Yet, it remains unclear what the exact shape of the latter specimen is, and whether it had barbs on the missing, more proximal part of the tool. All specimens show heavy traces of use. There are visible compaction marks and compacted tissue at the base of the AMSdated specimen 10016 ( Fig. 13:1), which in combination with the presence of a perforation close to the base of the same specimen may point to the use of this barbed point as a proper detachable harpoon head inserted into some sort of haft in order to maintain the link between the prey and the hunter. However, we cannot exclude its possible use as a projectile arrow or that it might have been used with a spearthrower. 56 Here we prefer to use a more neutral term barbed point (designed to obstruct easy extraction from a wound) rather than the commonly used designation "harpoon." The latter is considered as an operating mode rather than necessarily a particular morphological category of weaponry, i.e., it refers to the mode of predation when a detachable head is attached to a line, "drag," or impediment that always maintains the link between the prey and the hunter. 57 Based on his examination of the Upper Magdalenian barbed points against the known function of different morphologies of barbed weaponry among North American ethnographic cases, Pétillon 58 has shown that morphological features of barbed points (e.g., spurs, or holes on proximal ends of barbed projectiles), often used in making an argument that barbed points were used as detachable harpoon heads, are not su cient traits for unequivocal distinction between harpoons, which are primarily used in hunting sh, otter, beaver, and other water animals in known ethnographic examples, and a more general category of barbed projectiles, which could have be used as arrows when hunting large game or even as weapons of war.
With two direct AMS measurements (Table 2, Fig. 9) now available for two barbed points found during both research phases at Badanj, as well as judging by the stratigraphic position of other barbed specimens, it is possible to con rm that this type of tools is con ned to the Late period of the Badanj sequence. Specimen 10016, which was found in Level 6 and dated by OxA-2796-45, suggests the likely appearance of this technological innovation already from around 14 kya cal BP to 13.3 kya cal BP, a large span due to large error terms of this measurement. On the other hand, OxA-38111 dates another barbed point specimen (7263), found in Basler's trench but, unfortunately, with no precise contextual details. It gives a more precise measurement that falls into the last centuries of the 14th millennium cal BP (Table  2). Specimen 10037 (Fig. 12:3) was found in Level 2 and it would be expected that it is of even younger date, perhaps pushing its age to the end of the Bølling/Allerød interstadial in the rst centuries of the 13th millennium cal BP, or perhaps even later, but this can only be con rmed by future targeted AMS dating if not of this particular specimen then of datable materials from the uppermost levels of the site. For specimens 7264 and 7807, found during Basler's excavations, recorded depths of 0.5-0.6 and 0.6-0.7 suggest that these clearly belonged to the Late period of the stratigraphic sequence.
These new direct AMS measurements provide an unambiguous con rmation that currently the barbed points from Badanj represent the earliest directly dated appearance of this technological innovation in the  64 Based on two existing radiocarbon measurements from this layer, this specimen is later than the dated Badanj specimens (see Table 2). Another two pointed barbed points in the wider Eastern Adriatic region have recently been found in the Epigravettian levels of the site of Vlakno on the island of Dugi Otok (Fig. 2). 65 Different morphologies found among the Badanj barbed points may point to a long period during which some of these artifacts were sporadically used, as perhaps suggested by the two new radiocarbon measurements and the stratigraphic position of currently undated specimens. Yet, the morphological variability could also point to a long period of experimentation with this novel technological solution in modifying existing patterns of predatory hunting behavior at Badanj.
As to the function of barbed points at Badanj in targeting speci c prey, we could speculate that their appearance with the start of the Bølling/Allerød interstadial is not coincidental and that, similar to other cases in Europe, this could relate to the expansion of forest cover with the onset of warmer temperatures. It is also the case that at Badanj the appearance of barbed points from Level 6 onwards coincides with changes in the composition of lithic industry, which was now for the rst time characterized by the appearance of geometric microliths, such as lunates and triangles (see above), 66 which were likely used as armatures for composite tools, possibly forming "lithic barbs," perhaps with some use modi cations from the earlier use of backed bladelets in a similar manner. 67 The likely homology between the use of such composite tools and antler barbed points seems unavoidable, even though the use of antler barbed points seems to have been rather restricted at Badanj compared to the abundance of int armatures. The common assumption about the use of antler barbed points (through a harpoon operating mode) for shing and fowling cannot be excluded even though the specimens of both taxa exhibit rather restricted frequencies in the Badanj faunal assemblage (see above 68 ). Other possible small prey targeted by using barbed points might have been lagomorphs (hare, rabbit), which could have some support in the steady presence of this category of prey in the composition of the faunal assemblage throughout the Badanj sequence. 69 Finally, following the suggestion that based on the evidence for the introduction of entire carcasses of red deer to the site, 70 red deer might have been hunted in the vicinity of the site, with the Bregava River possibly serving as a corridor for the transit of the herds of red deer, we may speculate that antler barbed points might have also been used for hunting this and other ungulate species trapped in water. 71 Incised bone specimens Sequential and subparallel regular incisions are noted on four examined bone objects ( Fig. 15:1-3, 6).
Some or all of these specimens might represent broken parts of bone tools, likely utilizing long bone/metapodial diaphyses of a medium to large mammals. Incisions are relatively deep on three specimens and somewhat shallower incisions are found on one specimen ( Fig. 15:2). They were created by a to-and-fro grooving action. Most of the incisions are relatively short notches but in one case some of the present incisions appear longer ( Fig. 15:1). There does not seem to be a special preparation of the surface where these notches were made. While these objects are highly fragmented, which may prevent us from understanding the full extent of this sort of sequential incisions over the surfaces of the objects, only one set of incisions seems preserved on each tool apart from specimen 7602/1. We have tried to obtain AMS dates on two of the incised objects (10022 and 7602/1) but the samples were withdrawn due to low pretreatment collagen yield (Table 2). Yet, the stratigraphic position of all four objects can be con ned to the Late period of the Badanj sequence. One of the objects comes from Level 4 and was There are two bone objects from Badanj that differ from simple sequential incisions just described and exhibit more complex non gurative decorative patterns. Specimen 7540 (Fig. 15:4) is evenly graycolor burnt bone with old breaks. It exhibits a speci c pattern of decoration, which can only be properly seen aided by a microscopic magni cation ( Fig. 16:2). The engraved curvilinear line was probably incised rst by a burin-like tool leaving a very deep and wide groove, showing a heavily striated bottom. The curvature that the line makes is followed on both sides of the deep groove by a series of short and relatively shallow parallel to subparallel carefully executed incisions. On one, inner side of the groove where it forms an approximately 60-degree angle, the line consisting of diagonally (in relation to the main direction of the groove) placed incisions, is found immediately next to the groove. On the other, outer side of the deeply grooved line, at a short distance from the groove, three parallel curved lines are created by the same technique of placing short diagonal incisions. There is roughly equal spacing between these three lines and the deep groove.
Specimen 7541 (Fig. 15:5, Fig. 16:1) likely represents a proximal part of a broken tool made on a long bone/metapodial diaphysis of a medium-size mammal. The surface of this object was fully scraped, which might have created the unusually even beige/white coloration when compared to other bone specimens from the site. Subsequently, a multiple parallel chevron-like motif was incised on the outer surface of the shaft. The incisions are lled with a dark material, which might have been some sort of resin. This might have been done deliberately in order to create a visual contrast between the incised dark motif and the bright whitish background. Although the surface is completely covered by consolidants, we are currently pursuing further means of identifying what was used as the lling inlay for enhancing the visibility of the engraved motif.
Both of these decorated objects found during Basler's excavation at Badanj come from the Early period in the Badanj's stratigraphic sequence judging by their respective depths of 1.2-1.3 m (inv. 7540) and 1.5-1.6 m (inv. 7541). While the sample of these two objects is obviously small for far-reaching conclusions, it could be that more elaborate decoration is con ned to the early part of the Badanj sequence only. This could perhaps also be linked to the stratigraphic position of the engraved rock (see above), which was incised, according to the excavator, in the earlier parts of the sequence and covered by occupation sediments of the later Palaeolithic inhabitants of Badanj. 75 Engraving of diverse sets of decorative motifs on stone (even int) or osseous materials, including items of adornment (see also below), is common on Epigravettian sites in the Balkans and Italy. One particular incised stone from Romanelli is similar to specimen 7541 from Badanj in the diagonal execution of parallel incisions forming a line along the incised groove. Another such motif is found on a rock from Grotta del Cavallo. 76 Chevron-like motifs have been found in Grotta del Cavallo 77 and Riparo Villabruna. 78 On an ornamented piece of bone from Grotta Paglicci, chevron-like motif was incised over the incised gurative depiction of an ibex. 79 Deeply grooved curvilinear ornamental motifs are found in Early

Personal Ornaments
There is a large collection of items from the excavations at Badanj that most likely served as personal ornaments (over 1000 specimens from the 1976-1979 excavations alone according to the listings in the museum inventory book, of which we examined 877 specimens) comprising the following mammal body parts and mollusc species: bovid incisors (only two specimens which we will not discuss further), red deer (Cervus elaphus) canines, Tritia gibbosula, Tritia neritea, Columbella rustica, Dentalium sp., Glycymeris sp., Pecten maximus, Lithoglyphus naticoides, and Theodoxus danubialis. Non-edible marine and freshwater gastropods, scaphopods, and bivalves were used to create ornaments.

Red deer canines
We examined 31 specimens of red deer canines from Badanj (Fig. 17). The specimens are moderately well preserved. Most of this type of beads in the museum collection had a perforation on the root that was likely used for suspension. A lateral pattern of breakage is noticeable on a number of these holes ( Fig. 17:1, 3, 14), which would suggest that these were suspended, perhaps being sewn up, on one of the narrow lateral sides. On some others, the tip of the root was broken (Fig. 17:2, 5, 9). The holes were created on the root through the preparation of this zone by scraping, which resulted in the thinning of this area in order to facilitate easier perforation (Fig. 18). One specimen was probably intentionally burnt in order to achieve a uniform black color, and a highly polished and shiny surface (Fig. 17:1). A comparable case where the argument about deliberate thermal alteration of beads was made relates to Tritia neritea ornaments from Upper Palaeolithic and Mesolithic levels of Franchthi cave in Greece. 95 There are two specimens that are decorated by transversal incisions over their lobe surfaces (Fig. 18).
One of these specimens (7269) had transversal incisions that were very deep and were made using an unretouched int. An analogous example of transversal incisions over the lobe of the red deer canine is known from the Epigravettian site of Riparo Tagliente. 96 As red deer canine beads were found at different depths measured from the surface, it seems that they were used throughout the sequence, appearing in both Early and Late periods of the site occupation.
Red deer canines are a relatively common type of beads on both early and late Epigravettian sites in the Adriatic Basin (e.g., Šandalja II 97 ; Vela Spila 98 ). 99 We plan a more detailed study of specimens from Badanj by applying criteria for determining age and sex of individual specimens, and in determining whether any pairs coming from both lateral sides of the same individual can be recognized. 100

Tritia gibbosula
We examined 25 specimens of Tritia gibbosula found at Badanj (Fig. 19:1-8). Beads made of this marine gastropod are overall in a poor state of preservation. Also, some specimens exhibit taphonomic alterations, such as holes created by perforating organisms (Fig. 19:3, 8, Fig. 20:3). All specimens show anthropically made perforations. It is likely that direct pressure was applied from inside to the outside in order to create these holes. Ornaments show developed traces of use indicative of their prolonged handling. Holes are all very worn and enlarged ( Fig. 19:5, 7, 8, Fig. 20:2). Red ochre residues, when present, are sometimes scattered over the outside surface of the shell, which may indicate that these derive from the surrounding sediment. However, on some specimens, there are visible ochre residues along the rims, i.e., opening of the shell (Fig. 20:12), suggesting that these red ochre residues probably come from colored organic (sinew?) strings that might have been used to attach (embroider?) these beads to an organic surface. 101 Yet, it still remains di cult to distinguish the presence of residues from modes of suspension to residues that stick to a shell due to contact with the surrounding sediment.
Based on the depths associated with the studied specimens of Tritia gibbosula, all of the beads come from the Early period in the Badanj sequence. Apart from Badanj, among Epigravettian sites in the eastern Adriatic Basin catchment zone, Tritia gibbosula was reported to have been found at only two sites: seven specimens came from Vela Spila on the island of Korčula 102 and two specimens were found in the Epigravettian levels of the site of Mališina Stijena in the deeper hinterland of Montenegro, 103 the latter dated to ca. 17-16 kya cal BP ( Table 2). The absence of these beads in the late Epigravettian (after ca. 14 kya cal BP) seems to relate to a relatively con ned period of their cultural popularity in this regional context but further investigation into possible environmental reasons for this absence are necessary. Moreover, considering the wider patterning of evidence for body adornment across southeastern Europe throughout the Palaeolithic and Mesolithic, it is worth noting that Tritia gibbosula beads are found only within the Eastern Adriatic catchment zone. 104

Tritia neritea
There were 326 specimens of Tritia neritea in the collection we were able to examine (Fig. 19:10-29). A number of specimens exhibit post-depositional alterations, such as concretions from the sediment and sometimes water-related activity that exfoliated surfaces of some beads and contributed to their poor state of preservation. Some specimens are entirely burnt while some others are partially burnt so it remains di cult to determine sometimes whether burning was deliberate, as previously suggested for a red deer canine bead (see above), or certain specimens were accidentally burnt due to the burning of the surrounding sediment in which particular beads were deposited. Judging by the evenness of burning and high polish of the surface of a number of specimens one could make a good case for deliberate burning of at least a portion of the assemblage of Tritia neritea, which is similar to the treatment of the same bead type at Franchthi cave in Greece. 105 Perforations on the shells were made by gentle pressure from the inside to the outside, i.e., from the mouth of the gastropod outside. There are well developed traces of use and rounding on most of the perforated pieces, suggesting, as with other types of beads in this assemblage, a prolonged period of their use. Preserved residues from red ochre on these beads pose similar dilemma as in the case of Tritia gibbosula beads, but it is clear that a number of residues relate to the modalities of suspension. For instance, there is a clear pattern of red residues on the edge of the anthropically made perforation on one of the specimens (Fig. 20:7 -8). This would again suggest that likely red ochre colored organic (sinew?) strings might have been used to attach (embroider?) these beads to an organic surface. Another specimen shows a clearly visible red colored strip along the lateral side of the shell (Fig. 20:9), suggesting it was either colored intentionally or this strip was created through the contact of the shell and a colored material. Future experimental activity may better account for this particular pattern of residues.
When examining information on the depth measurements for the stratigraphic distribution of this type of beads at Badanj, they are found throughout the Badanj sequence. Tritia neritea beads are a widespread type of ornament throughout the Palaeolithic and Mesolithic of Mediterranean Europe. 106

Columbella rustica
We examined 59 specimens of Columbella rustica beads found in the Badanj collection ( Fig. 19:30-41). These specimens are slightly better preserved than other types of ornaments. Different from Tritia gibbosula and Tritia neritea beads, perforations on Columbella rustica were made through indirect percussion from the outside, possibly by using a int tool and a pebble. Traces of this procedure have left striations on the shell surface around the lower and upper part of the hole. Some of the holes are large with rounded rims of the perforation, suggesting a prolonged period of use. There are clear residue traces from a xing of the shell to an organic surface where the red ochre residues show up as a rim all around the pro le of the shells . A number of specimens are of evenly dark burnt coloration  and, as in the case of some red deer canines and Tritia neritea beads (see above), these specimens might have been deliberately burnt to achieve a desirable color of the bead.
Based on the depth measurements, it seems that Columbella rustica are con ned to the Late period in the Badanj sequence. This would correspond with other strands of data from other sites in the Adriatic Basin and Greece that suggest a shift towards popularity of Columbella rustica in the late phase of the Upper Palaeolithic and particularly in the Mesolithic. 107 Dentalium sp.
Based on the number of individual specimens, Dentalium sp. is the largest group in the assemblage of personal ornaments from Badanj, with 371 specimens that we have been able to examine .
Their overall state of preservation varies, but they are moderately well preserved. The applied consolidants over their surfaces prevent us from understanding well the way these specimens were cut, i.e., segmented. By and large, segments tend to have a standardized size of 2 to 3 cm, even though there are several very short and several longer specimens. Dentalium shells show rounding and beveled modi cations of the rims of these segments (Fig. 20:1), which is an indication that these ornaments were used for a long period of suspension or in contact with soft organic material. Many specimens are heavily worn.
Based on the depth measurements, the majority of these beads come from the levels 1.0 m and higher, with only a handful of specimens being found at the depths of 1.4 m or 1.3-1.2 m. This would suggest that the popularity in the use of Dentalium sp. peaks in the Late period of occupation, which corresponds well with stratigraphic observations about the presence of these beads in the levels excavated in 1986-1987. 108 Beads from Dentalium sp. were found in southeastern Europe since the Aurignacian period, and their popularity increases in the Late Upper Palaeolithic, but in this regional context Badanj stands out as the site with the highest frequency of Dentalium sp.
There were 53 whole or fragmented valves of Glycymeris sp. at Badanj (Fig. 21:1-13). Overall, the state of preservation of these specimens is poor due to taphonomic smoothing and exfoliated surfaces or occasional rounding. Thermal alteration of one of the pieces (Fig. 21:6) was probably accidental as it is of uneven coloration. Some specimens might not have been ornaments as they lack perforations, at least on the preserved part of the valve. But most of the best preserved specimens have clear perforations in the proximity of the umbone of the shell (Fig. 21:2 , 4, 6, 8, 11). The holes show developed use-wear traces, thus effacing technological traces. The use-wear traces suggest that these were possibly suspended as personal ornaments.
While the majority of specimens seem to be con ned to the upper part of the Badanj sequence, i.e., to the Late period, a fair number of specimens were found below 1 m of depth, i.e., down to a depth of 1.4-1.5 m, and this would suggest that this type of ornament was used throughout the Badanj sequence. One complete unperforated Glycymeris sp. valve and two fragmented pieces come from Epigravettian Layer where, among other things, they might have had the role of pigment containers. 111 Badanj here again stands out compared to other southeastern European Upper Palaeolithic sites with by far the highest number of Glycymeris sp. specimens.

Pecten maximus
Only a couple of fragmentary pieces of Pecten maximus (Fig. 21:14-15) were found. As these specimens do not exhibit perforations, we could not be certain these were used as ornaments at Badanj. However, specimens of this type of shell bearing perforations have been documented in other parts of the Upper Palaeolithic Mediterranean. 112 At Badanj, these specimens seem to be con ned to lower parts of the sequence.

Lithoglyphus naticoides
There were eight specimens of this freshwater gastropod at Badanj. Only four specimens had anthropically made perforations and could be considered beads. Due to the evenness of the color achieved by burning one of these specimens (7432/5, Fig. 19:9), we could suggest that it was deliberately burnt in order to create a desirable color of the bead, similar to other types of thermally altered beads in this assemblage (see above).

Theodoxus danubialis
In the museum collection, there was only one anthropically perforated specimen of freshwater gastropod Theodoxus danubialis.

Discussion
In hunter-gatherer studies of the Balkan Peninsula, one of the main theoretical models, proposed in the 1960s by 113 suggested that we should expect seasonal transhumance of human groups with the pattern of winter occupation in near-coastal areas and summer habitation in the uplands and generally hinterland regions of the Adriatic. This model was in particular suggested for the sites in Epirus, Greece. In a modi ed version of this model for the same region, Bailey et al. 114  Taking into account the location of Badanj against the background of the proposed models for forager logistical mobility and seasonality, the prediction would, then, be that groups occupying more permanent site locations in coastal areas during the winter months might have occupied Badanj en route to more upland areas where resources aggregated during the warmer months. These stays at Badanj, according to the model prediction, thus might have occurred either in the spring or fall. 120 Also, any changes in environmental predictability in the course of the Badanj sequence, for instance the start and the end of the Bølling/Allerød interstadial, would have created deviations from the regular pattern of use, not least because of changes in the availability of territories on the North Adriatic Plain due to the sea level rise, thus spatially constraining both the distribution of animal herds and human populations. Contrary to the model predictions, the existing seasonality indicators suggest a strong pattern of winter or late fall through to the early spring occupation of Badanj with some indicators possibly suggesting a shift in the seasonality of occupation in the later part of the sequence. While one has to be cautious with potentially unrepresentative small sample sizes for inferring ne temporal changes in seasonal scheduling at Badanj, it is clear that the site does not conform to the model for winter occupation of coastal areas and summer occupation of the hinterland. Badanj is here taken as a hinterland site despite being found in a lowland site location. 121 On the other hand, while the model of environmental stability and settlement redundancy could perhaps apply to the earlier parts of the sequence, with a break in the expected redundancy of the season of occupation after Level 6, we need both a much better hold over the chronological resolution of the Badanj sequence and more robust seasonality indicators so that changes in the seasonality of occupation and the timings of various environmental shifts can properly be correlated. Despite the strong indicators of the winter occupation of Badanj, Miracle and O'Brien 122 hesitate to rebrand Badanj as a base camp in a wider regional settlement system and conclude that the site might have been used as a special activity camp to hunt and process female red deer. At this point, apart from the focus of previous studies of Badanj on either the chipped stone industry or faunal remains as the most dominant classes of archaeological material, we believe that two categories of material culture examined in this article could potentially help in piecing together the puzzle of the site's (un)changing roles for forager groups of the region in a diachronic perspective.
Barbed points in the assemblage of osseous artifacts are of particular interest among other osseous tools as their appearance coincided with the start of the Bølling/Allerød interstadial, as shown by new AMS dates, as well as with several other changes, among which the most relevant might have been changes in the structure of lithic tools with the appearance of lunates and triangles as likely armatures for projectile points (even though this assumption needs further con rmation through dedicated use-wear studies). Such composite artifacts with "lithic barbs" might have conveyed the same idea to antler barbed projectile points in a different material. While genomic data still remain limited at present, recent analysis 123 suggests a widespread Villabruna cluster of genetic ancestry across Europe in the period after 14 kya cal BP, implying higher levels of mobility that might have gone in tandem with the phase of rapid warming across the continent. The tradition of Azilian osseous barbed points from the areas of southern France might have at this time reached Badanj, but it is also true that diverse cultural traits might have been transmitted in different directions. Further, the non-gural decorative patterns on some of the osseous artifacts from Badanj that we examined suggest a shared repertoire of patterns across wide areas of Italy and the Balkans at this time, further strengthening the idea of high levels of regional mobility and transfer of information. 124 On the other hand, we have already noted high frequencies and varieties of beads at Badanj when compared to other contemporaneous Epigravettian sites in the Adriatic Basin. As the excavation campaigns during which these beads were unearthed were made in the 1970s by the same excavator that previously worked at Crvena Stijena,125 where much smaller frequencies of beads have been found in contemporaneous levels 126 over an area larger than the one investigated at Badanj, it is unlikely that recovery bias can explain this situation. 127 Both sites are also at similar distances from the nearest coast of the Adriatic Sea with similarly rugged terrains in between. The contrast is further enhanced when Badanj is compared to coastal or near-coastal sites in the Adriatic (e.g., Vela Spila, Vlakno) where modern standards of excavation and recovery of archaeological materials with extensive programs of sieving were applied and which still produced smaller frequencies of ornamental beads. This patterning strongly suggests that although one might well predict that near-coastal locations will exhibit higher frequencies of marine species, Badanj as a hinterland location de es this prediction with a much higher percentage of different types of marine species used for ornaments than anywhere else (only Franchthi cave shows much higher frequencies of certain ornament types, but the presence of Glycymeris sp. at Badanj is signi cantly higher than at any of the extensively excavated levels at Franchthi). At the same time, these high frequencies of marine ornament types at Badanj do indicate very strong links with the coastal region, either through a pattern of logistical mobility of the groups using Badanj, that might have spent a part of the year in coastal areas where these ornaments were collected directly, or, alternatively, the ornaments might have been acquired through a network of exchanges of people and/or material culture.
Further interesting patterning regarding Badanj relates to the intra-site distribution of ornaments, where most of the material was found in the eastern-central area of the rock-shelter excavated by Basler, i.e., in the area around the engraved rock, rather than in the western zone where the 1986-1987 excavations took place. Badanj is unique in the presence of this engraved rock, 128 which is a rare occurrence of Upper Palaeolithic "art" in this regional context. 129 While we still need to obtain precise counts of all ornaments from Badanj from both phases of excavations along with more detailed analysis of spatial and vertical patterns of distribution of these ornaments, by and large we do not expect that the observed intra-site pattern will change. A hypothesis could be put forward that the higher frequency of ornaments found in Basler's trench could in some way be linked to the presence of the engraved rock, possibly related to heightened levels of symbolic expression in this particular location within the site. We cannot exclude the possibility that the site itself, along with a range of everyday activities (hunting, processing of carcasses, hide processing, int-knapping, armature preparation, etc.) was also a place imbued with ceremonial and ritual activities, or a place of gatherings of wider groups that shared information or held ceremonies here in the winter months, often wearing ceremonial costumes on which various ornaments might have been attached. This could further be strengthened if we take into account Kuhn and Stiner's 130 suggestion that beads should be seen as a form of information technology for broadcasting social information to wide audiences. It seems di cult to imagine that unusually high numbers of beads for the eastern Adriatic region, with a speci c intra-site spatial pattern, the presence of a rare example of parietal art, very high frequency of material culture items, and a seasonal pattern of winter occupation are all coincidental occurrences.
We would tentatively suggest that Badanj might have acted as a nexus of the maximal or regional band, a persistent place, 131 in an extended network characterized by the assumed hexagonal packing of spatial units consisting of minimal bands as characteristic of hunter-gatherer settlement systems crossculturally. 132 In the light of this evidence, there should be no reason to exclude the role of Badanj as a winter base camp, which might have had several specialized roles for the Epigravettian foragers of the wider regional zone. Future analysis of the available archaeological material from Badanj can further test this hypothesis. In addition, a better understanding of the nature of occupation at other contemporaneous sites in this micro-regional context can provide further insights into the nature of the regional forager settlement system during the Late Glacial period, providing a useful contextualization of the evidence from Badanj.

Conclusions
In this article we reviewed the existing strands of data about the Badanj sequence and presented our analyses of the previously largely unpublished assemblage of osseous tools, including newly obtained AMS dates on morphologically diagnostic curated barbed points, and the large extant collection of ornamental beads. By doing this, we attempted to re-address the relevance of this site for studies of the Upper Palaeolithic Epigravettian period in the Adriatic Basin and beyond. The unusually high frequency of personal ornaments at Badanj along with their intra-site patterning around the engraved rock suggest to us that the site might not have been used solely as a specialized hunting site for targeting pregnant female red deer herds, as proposed by previous studies, although targeting of pregnant female red deer did represent one of the important subsistence activities at Badanj. Several lines of evidence highlighted in this article suggest that we should perhaps consider Badanj as a persistent place, an important point for gatherings in the winter months, a base camp, which might have included the exchange of information and a range of ceremonial and/or ritual activities, which might have been inextricably linked to practices of hunting, over a considerable period of time. We suggest that the large amounts of different bead types found here compared to other contemporaneous sites in the Adriatic Basin and elsewhere in southeastern Europe as well as the presence of the engraved rock, and a peculiar concentration of ornaments around this rock, must all be taken into account along with other indicators of site function.
The summarized data and presented material have opened up our thinking toward further research agendas with regard to this outstanding sequence. One aspect of future research that remains of pressing importance is to provide a more robust chronological framework for the complete sequence.
Further insights could also be gained through use-wear and residue analysis of stone tools, including ground stones. A better understanding of formation processes would also be possible through micromorphological and sedimentological analyses. Future work on a more ne-grained resolution of site-speci c spatial and temporal patterns should be complemented by further regional surveys and excavations of chronologically overlapping sequences in this regional context. Such novel strands of data would aide previous and current efforts in putting together a more comprehensive picture of Badanj's signi cance for Late Upper Palaeolithic foragers as they navigated challenges brought about by changes of environmental conditions at the end of the Pleistocene.