STUDY OF THE AURIGNACIAN-GRAVETTIAN «TRANSITION» EAST OF THE CARPATHIANS: bLADELET PRODUCTION FEATURES FROM MITOC-MALU GALbEN (ROMANIA) AND MOLODOVA V (UKRAINE)

Although the Aurignacian to Gravettian «transition» represents a considerable cultural change, this issue still lacks an explanatory consensus and needs further researches. Data is quite unequal throughout Europe, as exemplified by the middle Prut and Dniestr area, which have remained little investigated so far, despite the presence of different sites related to Late Aurignacian and Early Gravettian. Some of these sites consist of secure sequences with high-resolution climatic context and abundant archaeology, such as Molodova V (Ukraine) or Mitoc-Malu Galben (Romania). Rich in lithic artefacts, these collections have rarely been examined else than typologically. In order to highlight this phenomenon in the concerned area, we have studied and confronted the bladelet production patterns from Mitoc-Malu Galben’s Late Aurignacian to Molodova V’s Early Gravettian. Preliminary results emphasize the technological individuality of both technocomplexes, but further investigations on materials and contextual data are necessary to confirm any population dynamics model.

Introduction. Aurignacian and Gravettian both represent major well-defined Upper Palaeo lithic cultures, associated to several respective specificities justifying their uniformity over time and space. It is consensually accepted that the appearance of the Aurignacian in Europe, before 40000 BP, is connected to the movements of Homo Sapiens (Hublin 2015;Nigst et al. 2014); despite some evolution, its material features stood stable enough to consider it a consistent cultural entity for more than 10000 years (eds. Bar-Yosef, Zilhгo 2006;eds. Zilhгo, d'Errico 2003;Kozłowski, Otte 2000). Starting from 30000 BP approximately, the cultural landscape changes with the first oc currences of the Gravettian. New behaviours and material features attest to this change (ed. Gou tas et al. 2011;ed. Otte 2013), while its material general characteristics relate to the Upper Pal aeolithic just as the Aurignacian.
The nature of this shift is still debated (Kozłowski 2015). On one hand, a simultaneity model stems from the observation that several early Gravettian sites throughout Europe bear dates reaching as far as 27000-28000 BP. Fac ing such facts, a general adaptative interpretation is relevant. But, on another hand, a monocentric model is also seriously considered, if not dominat ing. Based on the idea that Gravettian «innova tions» have first appeared in a specific area, such a model emphasizes the importance of popula tion movements or cultural diffusion. In this per spective, Central Europe is usually regarded as the most probable primary centre (Otte, Noiret 2004), with dates for the Gravettian going over 29000 BP, or maybe 30000 BP with less confi dence (Haesaerts et al. 1996, Kozłowski 2015). An inflexible point of view should preferably be left aside while considering those hypotheses, as both are defendable, and as several factors are prob ably involved. Insights from genetics, although preliminary, have suggested a real but unequal impact of population movements and replace ments (Fu et al. 2016), strengthening this as sertion. Facing a continent-scaled phenomenon, regionally differentiated explanations are, to say the least, cautiously justified.
Concerning this issue, Eastern Europe is also involved, even though the situation is unclear and complexified by the recognition of other cul tural entities (e. g. Streletskian, Gorodtsovian) in the same time slot (Sinitsyn 2015). Specifically, the area encompassing the middle Prut and mid dle Dniestr basins is no exception. Both cultural entities are present there, but chronologies often lack precision (Noiret 2009). Indeed, assured Au rignacian sequences are scarce and often approxi mately set chronologically. If sites further East such as Kostienki 1/III (Sinitsyn 1993(Sinitsyn , 2015, Ko stienki 14/LVA (Sinitsyn 2003(Sinitsyn , 2015 or Siuren I (eds. Demidenko et al. 2012) support the recog nition as Aurignacian, in the whole Moldavian area Mitoc-Malu Galben stands as the only reli able occurrence. The site of Corpaci-Mвs, also on the Prut river, might be the only other Aurigna cian instance, but published data do not allow to go over uncertainties in dates and attributions of this assemblage (Noiret 2004;2009). Despite this bad representation, the data suggests that an archaeological bias, e. g. representativity or recognition, could be considered explanatorily; a functional bias has also been proposed (Hoffecker 2011;Hoffecker et al. 2018). The picture is dif ferent for the Gravettian, as instances related to this wide technocomplex are numerous (Noiret 2007;2009). The early occurrences, however, are almost exclusive to the Molodova V sequence, which stands as a cornerstone in the local model. Its early Gravettian horizons (labelled 10 and 9) reach 29000 BP or even earlier for level 10 (Hae saerts et al. 2003). Unfortunately, published ma terials and associated data are often limited to typological approaches, unfit to characterize such phenomenon, and requiring new investigations.
Material and methods. In the lithic domain, Aurignacian and Gravettian materials logically show common features, both being Upper Palaeo lithic cultures. Indeed, despite the developments of technological approaches, the distinction be tween both technocomplexes still partly relies on the recognition of typological features to assess with certainty. Nevertheless, distinctive techno logical features are a reality, even though they are often subject to variability, due to the wide ness of these entities and the impacts of non-cul tural constraints. This said, the bladelet produc tion domain is quite differentiated between both technocomplexes. More especially, the Aurigna cian bladelet production is often considered to bear particular technological features, already crystallized under different typological forms (e. g. carinated burin and endscraper) for decades (ed. Le Brun-Ricalens et al. 2005;Le Brun-Ri calens 2005). Concerning the Gravettian bladelet technology, focus was often set on the products themselves, highlighting more peculiarities than general characteristics (Pesesse 2013). Still, the observation of Gravettian blanks' straightness and lightness is usually accepted. Finally, an ad ditional reason to investigate on bladelet produc tions before other aspects comes from their uses, frequently related to the projectile point domain. That kind of artefacts being highly typologized, their related technologies also have the potential to act as proxies in such an issue.
In the frame of this study, all artefacts related to bladelet production have then been examined following a technological approach, including re duction sequence reconstruction, refits and at tribute analysis; however, hereinafter presented data is mostly qualitative. Studied materials comprise primary and secondary bladelets, all as sociated maintenance blanks, and cores.
Focusing on the Prut and Dniestr area, atten tion was set on the most reliable and relevant sequences to acquire data: Mitoc-Malu Galben (MMG; Romania) and Molodova V (Ukraine).
MMG is one of the Upper Palaeolithic key-sites of Romania. Situated on the North-Eastern bor der of the country, along the river Prut, it consists of a 14 m high loess sequence, with multiple Au rignacian and Gravettian horizons (ed. Otte et al. 2007a). The modern archaeological investigations took place almost continuously from 1978 to 1995 (ed. Otte et al. 2007a), and from 2013 to 2016 (Li bois et al. 2018;Noiret et al. 2016). Human occu pations are divided in five Aurignacian and five Gravettian concentrations, labelled numerically from bottom to top («Aurignacian I», «II», «III», «III Supérieur»; Gravettian «I», «II», «III», «IV»); lowermost and uppermost layers are respectively labelled «Aurignacian Inférieur» and «Gravettian Indifférencié». A large radiocarbon dating cam paign led in a stratigraphically controlled envi ronment allowed to precisely and consistently set all archaeological horizons (Haesaerts 2007). In MMG, the Aurignacian spans then from 33000 to 27500 BP, while the main Gravettian occupa tions are set in the 27000-23000 BP time slot. Particularly, the first consistent horizon, «Aurig nacian I», is dated around 31000 BP, while later Aurignacian levels, «III» and «III Supérieur», are respectively dated from around 29500 BP and at 27500 BP. Materials originating from the 1978-1990 excavations were unfortunately subjected to selection; especially tools, cores, and some pri mary blanks were preserved. Later excavations happily provided with complete assemblages, still mostly unsieved in the 1990 s (Noiret 2005, p. 448).
From this site, two representative complete samples were taken into account for this study, coming from horizons «Aurignacian I» and «III». Moreover, further details from «Aurignacian III Supérieur» level were obtained by examining tools and cores from the older phase of excava tion (1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990). In the frame of this project, no Gravettian bladelet assemblages from MMG have not been examined yet. The chosen «Aurigna cian I» sample corresponds to the lithic remains associated to a hearth sampled in squares L5-L6 by the Belgian team in the 1990 s (Noiret et al. 2006). The considered «Aurignacian III» assem blage originates from the 1990 excavation. Limit ed to squares F03 and G03, this is one of the only complete assemblages of the 1978-1990 phase of excavations.
On the other hand, collections from the re nowned site of Molodova V were partly studied.
Situated in South-Western Ukraine, in the mid dle Dniestr basin, this 25 m high multi-layered sequence was excavated from 1951 to 1964 (eds. Ivanova, Tzeitlin 1987). It encompasses numerous Middle and Upper Palaeolithic horizons, mostly related to Mousterian, Gravettian, Epigravet tian and Mesolithic; layers 10 and 9 are the first Gravettian ones. The sequence was chronological ly set with rather good precision by the excavating team (Ivanova 1987); later, a revision campaign also deepened the absolute chronology resolution (Haesaerts et al. 2003, Haesaerts 2007. Even though layer 10 isn't precisely dated, layer 9 is radiocarbon dated at the latest around 29000 BP. As far as known, fieldwork didn't include sieving, but all materials were kept after excavation; so far, storage issues only are incriminated for the loss of materials. Most of the materials related to archaeologi cal horizons 10 and 9 were excavated in 1953 and 1954. A limited number of pieces comes from years 1955, 1960 and 1962; for a question of assemblage consistency, those were not taken into account in this frame. Layer 10 is complete at about 70 % in comparison to the published counts (Chernysh 1987). Unfortunately, missing materi als include part of the tools and most of the blades and bladelets. Bladelet production details were then inferred from cores and a limited number of products. Spatial analysis, based on markings, has revealed two main clusters of artefacts; due to their high representativity in bladelet production related artefacts, pieces from cluster 2 only were included in this study (squares 12 to 14, and E to K). Layer 9 presents less lacks, as almost 90 % of the materials were recovered; there doesn't seem to be any strong biases in artefacts classes fre quencies. No spatial-based selection was applied in the frame of this study to layer 9 materials.
Results. Aurignacian bladelet production schemes are pretty-well illustrated by results from the complete samples examined in MMG's «Aurignacian I» and «III» levels. The «Aurigna cian I» level was already well documented, as it is one of the most prolific in the MMG sequence. The older phase of excavations had already assured with the existence of some bladelet production in this level through the presence of numerous carinated tools and associated typological forms (Otte et al. 2007b). Only the detailed goals of the production remain unknown, still now, due to the lack of retouched bladelets from any phase of the excavations. Nevertheless, the L5-L6 sample gives new information on the progression and characteristics of the bladelet production, based on almost 200 bladelet elements and several re lated products. Among noticeable features is the high proportion of twisted shaped bladelets and naturally pointed ones ( fig. 1: 1); most of these blanks do not result from shaping or maintenance phases but from main production. Some mainte nance products are also clearly identifiable, as narrow tablets ( fig. 1: 2) and lateral maintenance flakes ( fig. 1: 3), the latter displaying lateralized bladelet removals on their dorsal surfaces. Those elements attest the narrowness of cores and later al accentuation of convexities during production. Four sets of refits were reconstructed and support this scheme. More especially, sets 1 and 2 dem onstrate the imbrication of varied-sized bladelets. Set 1 proves the narrowing through massive lat eral removals ( fig. 1: 4), while set 2 clearly shows the symmetric and semi-rotating progression of production ( fig. 1: 5). All those features can clear ly be compared to «classic» carinated tools / cores production (ed. Le . Contrarily to level «I», the bladelet production from «Aurignacian III» horizon had never been deeply documented. Rich of about 60 bladelets and 12 cores, the F03-G03 assemblage is excep tionally well conserved for this level. Variability in bladelet sizes and shapes is important, as this set includes equally main and side products, as well as initialisation blanks ( fig. 1: 6); still, twist ed profiles are dominating. Several refit series could be realised on bladelets. Set 1 illustrates the succession of heavily twisted blanks envelop ing the production surface ( fig. 1: 7). Set 2 shows the removal of a hinged scar followed by two curved and twisted bladelets ( fig. 1: 8). As well, the refitting of some products on a core demon strates pretty well the progression of knapping on the side part of the core, to maintain the lateral convexity ( fig. 1: 9). In a general way, other cores from this sample express the same tendencies, expressing again a convergence of lateral and frontal surfaces ( fig. 2: 1, 2), feature at the roots of the bladelets' shapes.
Concerning the last Aurignacian horizon in MMG, «Aurignacian III Supérieur», data is re grettably limited to the few carinated tools / cores from the first phase of excavations. Unlike level «III», it reveals changes in proportions and progression of knapping, as narrowing is less in tense and production surfaces shorter ( fig. 2: 3,  4); a decrease of twisted shapes is expected for produced blanks. Despite these modalities' modi fications, the general scheme of production still corresponds to the Aurignacian; yet it is difficult to extrapolate the transformations' conditions due to the paucity of materials.
Regarding Molodova V/10 s bladelet production, detailed data comes only from two classes of ar tefacts: retouched bladelets and cores; no other artefacts could bring substantial information. Retouched bladelets consist of four elements, of diverse raw materials ( fig. 2: 5). If complete ele ments are standardized a bit over 30 mm, width and thickness are more varied. The importance of the back depends on blank's original width, as lighter blanks are less intensively retouched. Bladelets are straight or, at the most, barely curved. Data from cores is consistent with those Libois, T. Study of the Aurignacian-Gravettian «Transition» East of the Carpathians: Bladelet Production...
On the contrary, cultural layer 9 shows abun dant remains of unretouched bladelets and a few cores. Blanks' variability in size is important; the lack of microbladelets is probably due to excava tion methods. Profiles are dominated by straight and slightly curved blanks, with parallel sides, while distal terminations are varied ( fig. 2: 9). The twelve bladelet cores are mostly unipolar, al though some are exploited from two independent platforms ( fig. 2: 8). Production doesn't take place on large faces, but several pieces demonstrate the use of less narrowed blocs than in level 10. Still, cores are not subject to distal convergence, with prevailing squared sections. Additionally, a specific reduction modality based on burin-like technologies is supposed to produce small blades and bladelets, as exemplified by the presence of burins multiple ( fig. 2: 11). Some straight blanks can be connected to it ( fig. 2: 10).
Discussion. The origin of the Gravettian East of the Carpathians has already been discussed by several researchers, but proposed models usually have aimed at finding relationships with earlier cultural entities. Limited by the available data, they could only count on typological descriptions and few reliable absolute dates. It is in such a con text that unknown transitional industries or even the Szeletian (Anikovich 1992) were suggested as origins of the Gravettian. At last, the inter nationalisation of results from MMG allowed to consider the Aurignacian a relevant predecessor to the Gravettian in the Prut and Dniestr area (Borziak, Koulakovska 1998).
The Aurignacian presence in MMG was al ready well assured by former studies, but these new detailed results contribute to several issues. It appears that bladelet production modalities from horizons «Aurignacian I» and «III» are in details comparable to what is seen in the Au rignacian sphere throughout Europe (e. g. ed. Demidenko et al. 2012;ed. Le Brun-Ricalens et al. 2005;Nigst 2012;Sinitsyn 2015;Teyssandier, Liolios 2003). Also, the recurrence of occupations in MMG suggests a persistent Aurignacian pres ence in the area. The question of a particularly late occurrence of the Aurignacian relies mostly on the «Aurignacian III Supérieur» horizon. Even though lithic data is a bit ambiguous, the exist ence of a gap between this horizon and the next Gravettian ones strengthens its association to the previous levels. All this suggests the Aurignacian lasted until approximately 27500 BP on the mid dle Prut river.
On the Gravettian side, abundant materials from Molodova V/9 allow to define precisely the means and goals of the bladelet production. Cores do not show any convergence of their surfaces lat erally or frontally like in the Aurignacian. In fact, raw materials are especially chosen to install a straighter production surface with parallel sides, as reflected on the obtained blanks. Despite its limited status, data from Molodova V/10 cor responds to the same scheme and is irrefutably Gravettian. An Aurignacian interpretation of materials, relying mostly on the presence of a «carinated» endscraper, had already been pro posed (Allsworth-Jones 1986; Hoffecker 2011), but should be totally refuted. The last disruptive element lies in the presence of the famous bifa cial «point» (Chernysh 1987, p. 28, fig. 15), which is not explained. No bifacial shaping artefacts or related are present in the materials, and this tool should then be assumed as an import on the site. The lack of detailed contextual data does not al low to confirm its strict association to the rest of the assemblage, especially as no other site with clear context and equivalent chronological situ ation presents the same association of materials. Consequently, it should be considered that the Gravettian in Molodova V appears fully charac teristic and independent from the Aurignacian in the bladelet production domain.
Since Haesaerts' geological investigations and Noiret's synthetic works, the MMG and Molo dova V sequences have appeared to be the only ones able to highlight this «transition» phenom enon and build a consistent model. Strict obser vation of radiocarbon dates suggests that a case of contemporaneity is exceptionally documented, with an Aurignacian presence on the Prut while Gravettian features have already appeared on the Dniestr. However, this unique early Gravet tian occurrence is followed by a 2000 years old hiatus before its next observation, making it a very lonely instance. Indeed, a «classic» succes sion model should not be directly left aside, as the raw technological data from MMG and Molo dova V doesn't suggest any clear link between both industries. Facing such fact, validation of a model is only possible after having re-examined contextual data from Molodova V, to confirm the association of materials labelled as levels 10 and 9 with the stratigraphy and dates.
Conclusion.Mitoc-Malu Galben and Molo dova V find no analogies in the middle Prut and Dniestr area to question the Aurignacian to Gravettian shift. If the succession from one to the other makes no doubt, the modalities are still un known. The new technological data here acquired highlights the discrepancy or, at least, the lack of clear link in the lithic domain. If this allows to rule out any «transitional» hypotheses, no popu lation dynamics model can be confirmed. Simul taneity and strict succession scenarios both need further information from sites contexts. More es pecially, association of materials to stratigraphy should be investigated in the Early Gravettian of Molodova V.
Acknowledgment. First thanks go to Ruslan Koropetskyi in Lviv and Cristina Cordoş in Iaşi for both giving me complete access to materials studied and exposed in this paper. An additional greeting is meant for Larissa Koulakovska, for proposing me to participate in this conference, and hence to diffuse those preliminary results among Ukrainian colleagues.