Documenta Praehistorica XLIV (2017) Modelling the earliest north-western dispersal of Mediterranean Impressed Wares: new dates and Bayesian chronological model Didier Binder1*, Philippe Lanos2, Lucia Angeli3, Louise Gomart1, Jean Guilaine4, Claire Manen5, Roberto Maggi6, Italo M. Muntoni7, Chiara Panelli1,8, Giovanna Radi3, Carlo Tozzi3, Daniele Arobba9, Janet Battentier1, Mario Brandaglia10, Laurent Bouby11, François Briois5, Alain Carré1, Claire Delhon1, Lionel Gourichon1, Philippe Marinval13, Renato Nisbet12, Stefano Rossi14, Peter Rowley-Conwy15 and Stéphanie Thiébault16 1 Université Côte d'Azur, CNRS, Cultures et environnements. Préhistoire, Antiquité, Moyen Âge, Nice, FR didier.binder@cepam.cnrs.fr *Corresponding author< louise.gomart@cepam.cnrs.fr< chiara.panelli@gmail.com< janet.battentier@cepam.cnrs.fr< alain.carre@cepam.cnrs.fr< claire.delhon@cepam.cnrs.fr< lionel.gourichon@cepam.cnrs.fr 2 Université Bordeaux Montaigne, CNRS, Institut de Recherche sur les Archéomatériaux, Bordeaux and Université Rennes 1, CNRS, Geosciences-Rennes, Rennes, FR philippe.lanos@univ-rennes1.fr 3 Università degli Studi di Pisa, Dipartimento di Civiltà e Forme del Sapere, Pisa, IT luciaangeli78@yahoo.it< giovanna.radi@unipi.it< tozzi@arch.unipi.it 4 Collège de France, Paris, FR jguilaine@wanadoo.fr 5 Université Toulouse, CNRS, EHESS, MCC, Travaux de Recherches Archéologiques sur les Cultures, les Espaces et les Sociétés, Toulouse, FR claire.manen@univ-tlse2.fr< francois.briois@ehess.fr 6 Istituto Internazionale di Studi Liguri, Chiavari, IT romaggi2003@libero.it 7 Soprintendenza Archeologia, Belle Arti e Paesaggio per le Province di Barletta-Andria-Trani et Foggia, Foggia, IT italomaria.muntoni@beniculturali.it 8 Università degli studi di Genova, Dipartimento di Antichità, Filosofia e Storia, Genova, IT chiara.panelli@gmail.com g Museo Archeologico del Finale, Finale Ligure, IT arobba@museoarcheofinale.it 10 Università degli studi di Firenze, Studi per l'Ecologia del Quaternario, Firenze, IT mariobrandaglia@hotmail.com 11 Université Montpellier, CNRS, IRD, EPHE, Institut des Sciences de l'Evolution de Montpellier, Montpellier, FR laurent.bouby@umontpellier.fr 12 Università Ca' Foscari, Dipartimento di Studi sull'Asia e sull'Africa Mediterranea, Venezia, IT renisbet@tin.it 13 Université Montpellier, CNRS, MCC, Archéologie des Sociétés Méditerranéennes, Lattes, FR philippe.marinval@cnrs.fr 14 Soprintendenza Archeologia, Belle Arti e Paesaggio per la città metropolitana di Genova e le province di Imperia, La Spezia e Savona, Genova, IT stefano.rossi@beniculturali.it 15 Durham University, Durham, UK p.a.rowley-conwy@durham.ac.uk 16 Muséum National d'Histoire Naturelle, CNRS, Archéozoologie, Archéobotanique: Sociétés, Pratiques et Environnements, Paris, FR stephanie.thiebault@mnhn.fr 54 DOI> i0.43i2\dp.44.4 Modelling the earliest north-western dispersal of Mediterranean Impressed Wares: new dates and Bayesian chronological model ABSTRACT - The authors attempt to specify the diffusion pattern of the Impressed-Ware Neolithic (Im-presso-cardial complex, ICC), from south-eastern Italy onto the French Mediterranean coasts. Using ChronoModel® software, a Bayesian model was built with sets of dates obtained on well-contextu-alised, short-lived samples. The results highlight a clear tightening of the chronology in the so-called nuclear area (Apulia, Basilicata, Calabria) and a pioneer dispersal at record speed in the Tyrrhenian Basin. Moreover, they question the origins and initial developments of the Impressed-Wares techno-complex. KEY WORDS - Italy; France; Neolithic; Impressed wares; Bayesian modelling Modeliranje najzgodnejše severno-zahodne širitve posod tipa Impresso v Sredozemlju: novi datumi in Bayesov kronološki model IZVLEČEK - Avtorji poskušajo definirati vzorec difuzije neolitskih posod tipa Impresso (kompleks Impresso-cardium, ang. kr. ICC) iz jugovzhodne Italije do sredozemske obale Francije. Bayesovo modeliranje je bilo postavljeno s pomočjo programske opreme ChronoModel® na podlagi serije datumov kratkoživih vzorcev iz dobro definiranih kontekstov. Rezultati kažejo jasno oženje kronologije na t. i. jedrnih območjih (Apulija, Bazilikata, Kalabrija) in pionirsko širitev z rekordno hitrostjo v Tirenskem morju. Avtorji prav tako opozorijo na težave pri interpretaciji začetkov in prvotnega razvoja tehno-kompleksa posod tipa Impresso. KLJUČNE BESEDE - Italija; Francija; neolitik; posode tipa Impresso; Bayesovo modeliranje Issues The western dispersal Western Mediterranean Neolithisation processes are thought to have been supported by a polythetic complex defined as Impressed Wares or the Impresso-car-dial complex (ICC). Its precise sources are still controversial, despite genetic evidence concerning animals, cereals as well as humans that now clearly inscribe it in the framework of a peopling movement, originating from the Eastern Mediterranean (Hofma-novd 2016). Indeed, in terms of intensity, rhythms, routes and recombination, the scenario of this dispersal is still poorly known (Binder, Guilaine 1999; Manen 2014). ICC provides huge internal diversity within, for instance, pottery styles, which makes the visibility of interaction and evolution processes rather fuzzy (Guilaine 2003). Moreover, towards the West, the Eastern-Mediterranean Neolithic package lost many of its attributes, especially in the symbolic range, and the farther West we go, the greater is this loss (Binder et al. 2014). This addresses the problem of a cultural or social drift and admixture that could be at the origin of such a declension of the Neolithic paradigm in the Western Mediterranean. What could be at the origin of such apparent variability? On the one hand, as shown by its typical coastal impact, the north-western Mediterranean colonisation by Neolithic people was certainly based on maritime seafaring (Bernabo Brea 1950) or voyaging (Ammerman 2013), which is evidenced by the diffusion of the whole set of western Mediterranean obsidian sources, a diffusion that did not exist at all in the same area before the farmers' dispersal (Ammerman, Polgase 1997; Binder et al. 2012; Muntoni 2012; Pessina, Radi 2006; Tykot et al. 2013), and whose speed has been several times discussed (Isern et al. 2017; Zilhâo 2001). In addition, such movements could have been more or less erratic, and even from diverse origins. On the other hand, the evidence of the Late Mesolithic setting in a large part of the Franco-Italian area during the 7th millennium as well as a few Mesolithic radiocarbon dates from the first half of the 6th millennium (e.g., in the Middle Rhône valley and in the Tosco-Emilian Apennine) could support the coexistence of groups of early farmers and late hunter-gatherers, which could be at the origin of several scenarios of admixture, with consequences in terms of cultural mosaic and biological diversity (Binder 2013; Binder et al. 2017; Franco 2011; Perrin, Binder 2014; Marchand, Perrin 2017). Building reliable datasets and chronicles Comprehensive data are required to make this story clearer, including reliable radiocarbon chronicles from clear contexts, which is the topic of this paper. 55 D. Binder, P. Lanos, L. Angeli, L. Gomart, J. Guilaine, C. Manen, R. Maggi, I. M. Muntoni, C. Panelli, G. Radi, C. Tozzi, D. Arobba Building reliable and accurate chronicles depends on the rules governing proof, the capacity to run a quality approach, which means outlining some basic and generally well-known premises for dating, i.e. to use only short-lived samples (abridged SLS) and to guarantee their association with well-defined events. The earliest ICC settlements dated within such conditions are localised in Corfu (n = 1), in southern (n = 8) and central Italy (n = 3), in the Tuscan archipelago (n = 1), in Liguria and eastern Provence (n = 4), and Mediterranean Languedoc (n = 2). As a comparison, Dalmatia also provided nine ICC sites with reliable SLS-dates (Forenbaher et al. 2013; Mc-Clure et al. 2014; Produg et al. 2014). The south-eastern and central Italian settlements considered here are significant because of the stylistic variations of Impressa pottery, following Santo Tine's periodisation (Cipolloni Sampo et al. 1999; Grifoni Cremonesi, Radi 1999; Tine 1987): archaic, with rather disorganised impressed decoration; evolved or Guadone, with structured impressed decoration; recent, with lgraffita dentellata' and painted Lagano-da-Piede wares; a final phase, with Matera engraved and Masseria La Quercia painted styles. For this study, we did not take into consideration the late dispersal towards the North along the Adriatic coast, and we omitted late SLS-dates from Molise (Monte Maulo) and Abruzzo (Fonti Rossi) (Skeates 1994). Likewise, among the southern Italian set of sites, late dates from Masseria Santa-Tecchia and Masseria Candelaro which pertain to the development of Red Painted Figulina Wares, mainly in the second half of the 6th millennium BCE (Skeates 1994), are not to be taken into account in our topic, which focuses on the Impressa aspects predating the mid-6th millennium BCE. In Provence and Mediterranean Languedoc, ICC settlements display diverse pottery styles, marked by the importance of stab-and-drag, instrumental and pinched decoration, compared to the occurrence of shell impression, especially Cardium, which is very variable depending on the sites (approx. 50% at Pont de Roque-Haute as opposed to almost zero at Pendimoun). These Impressa aspects definitely predate the Franco-Iberian Cardial, a phase which develops during the second half of the 6th millennium, and even mainly the last third of it (Binder et al. 2017). Data from the major western Mediterranean islands is lacking. A significant number of Pre-Stentinello and Early Stentinello settlements, including Kronio style, is attested in Sicily, but with no SLS-dates published. No evidence of Impressa pottery sets is as yet provided in Sardinia, despite very early evidence of obsidian trade, such as at Peiro Signado (Briois et al. 2009) and Arene Candide (Ammerman, Polgase 1997). Only one early date made on charcoal is known from Su Carroppu (Luglie 2014), with uncertainties concerning the cultural attribution of the few sherds discovered. On the basis of pottery typology, a few Impressa indices, probably predating the Tyrrhenian Cardial, are now known in Corsica at Cam-pu Stefanu (Cesari et al. 2014) and at the Albertini rock-shelter (Binder, Nonza-Micaelli 2016 andforth-coming) but not dated yet. Data are critically lacking in Tuscany, Lazio and Campania. It has been thought that Late Mesolithic groups in the Tosco-Emilian and Ligurian Apennine were still settled during a part of the ICC dispersal and its early development; this is based on a few radiocarbon dates obtained from Castelnovian deposits, e.g., at Lama Lite 2, US6, c. 5770-5640 cal BCE (6620+80 BP, Rome-394, charcoal) (Castelletti et al. 1994; Dini, Fioravanti 2011) and Monte Frignone 2, US2, c. 5630-5480 cal BCE (6624+45 BP, LTL-2656A, charcoal) (Dini, Fioravanti 2011). In southern and central Tuscany, with the exception of new dates from Giglio Island discussed below, the available dates concern Cardial aspects from the Tuscan Archipelago (Tozzi, Weiss 2001) which are rather late in the perspective of this paper; in addition, no SLS-date are available for these contexts. In Lazio, the lacustrine settlement of La Marmotta displayed different early pottery assemblages, including Impressa, whose clustering is debatable, and which did not provide SLS-dates (Fugazzola Delpino, Pessina 1999). The situation in Campania is very challenging, with very early dates associated with diverse Impressa aspects probably connected to Apulia via the Ofanto valley, e.g., at Baselice (Langella et al. 2003) and at La Starza d'Ariano Irpino (Albore Livadie 2002); unfortunately, the associated dates are suspected of suffering from the old-wood effect. In total, a set of 17 sites offering 94 AMS dates (Fig. 1, Tab. 1) is considered here for a first appraisal of the earliest Neolithic dispersal from southern Italy and Corfu towards the North-West, using Bayesian modelling. Among these sites, when available, SLS- 56 Modelling the earliest north-western dispersal of Mediterranean Impressed Wares: new dates and Bayesian chronological model dates from the period immediately following (i.e. Red Painted Wares and Cardial) were used to provide constraints on the probabilities of the earlier dates. Controversial SLS-dates from Mesolithic Ter-ragne (Gorgoglione et al. 1995) are also discussed infra. To build this database, all material potentially affected by the marine or fresh-water reservoir effect, i.e. shells, was rejected. We kept the few (n = 5) directly dated human remains available from Pendi-moun (early Cardial levels; Le Bras-Goude et al. 2006) and from Arma dell'Aquila (Impressa levels; Biagi, Starnini 2016), since stable isotopes indicate that none of them were questionable. Collagen samples from animal bones (n = 14), including mostly domestic remnants, are also considered with a small risk, as the feeding of livestock with marine products (Balasse et al. 2005) was never evidenced in the region (Le Bras-Goude et al. 2006). Considering charcoal, we have selected 10 dates, after rejecting long-lived tree samples (e.g., oak or juniper) whose anatomical characterisation (e.g., twigs or last rings samples) was not explicit. However, we kept, with question marks, a series of six unspecified charcoal samples from Trasanello due to the close connection of this site with Trasano. In addition, one date from Monochrome Sidari obtained on oak charcoal was used to constrain successive impressa deposits at this site. We also considered 58 dates from charred fruits and seeds, and primarily cereals, including husk. We used Bayesian statistics through ChronoModel® software, version 1.5 (Lanos, Philippe 2015a; 2015b; Lanos et al. 2015) and IntCal13 14C curve (Reimer et al. 2013) to model each site separately. Similar methods have been classically applied using OxCal, for example, for the whole of Northern Italy (Pearce 2013) with consistent results at the general level, albeit with limited relevance due to the random intrinsic quality and stratigraphic accuracy of the samples. ChronoModel software is based on the Bayesian event date model, which is aimed at estimating the date of a target event (Dean 1978) from the combination of individual dates derived from relevant dated events. This model has a hierarchical structure, which makes it possible to distinguish between a target event date (any date of interest for the archaeologist) and dates of events (artefacts) dated by chronometric methods, typo-chronology or historical documents. One assumes that these artefacts are all contemporaneous, which is relevant to the date of the target event. On the other hand, the dates can be affected by irreducible errors, hence the possible presence of outliers. To take into account these errors, the discrepancy between the chronometric dates and the target date is modelled by an individual variance, which allows the model to be robust to outliers, in the sense that individual variances act as outlier penalisation (Lanos, Philippe 2015a; 2015b). Thanks to this modelling, it is not necessary to discard outliers because the posterior (in the Ba- Fig. 1. Earliest Central and Western Mediterranean Impressed Wares sites dated with short-lived samples: 1 Portiragnes - Peiro Signado; 2 Portiragnes - Pont de Roque-Haute; 3 Castellar - Pen-dimoun; 4 Finale-Ligure -Arene Candide; 5 Finale Ligure - San Sebastiano di Per-ti; 6 Orco Feglino - Arma dell'Aquila; 7 Giglio Island -Le Secche; 8 Ortucchio - Colle Santo Stefano; 9 Lucera -Ripatetta; 10 Manfredonia -Coppa Nevigata; 11 Molfetta - Pulo; 12 Matera - Trasano; 13 Matera - Trasanello; 14 Ostuni - Sant'Angelo; 15 Co-rigliano Catabro - Favella della Corte; 16 Vizula - Medulin; 17 Liznjan 20 Drnis ■VÍA- . & ■ _ □ 6 □ 4-5 3 161® „ 22 19DJ 24 2 □ 7 » Q 8 n □ 910 Olí 12-13 14 tíMm 3oogle Earth Kargadur; 18 Losinj Island - Vela Spilja; 19 Rasinovac; Pokrovnik; 21 Sibenik - Konjevrate; 22 Mirlovic - Skarin Samograd; 23 Zemunica; 24 Orebic Nakovana; 25 Kakanj - Obre; 26 Susac Island; 27 Corfu - Sidari. 57 D. Binder, P. Lanos, L. Angeli, L. Gomart, J. Guilaine, C. Manen, R. Maggi, I. M. Muntoni, C. Panelli, G. Radi, C. Tozzi, D. Arobba yesian sense) high values of the individual variances will automatically penalise their contributions to the estimate of event date. Stratigraphic constraints, when available, are applied between event dates. Note that there is often only one chronometric date per target event date. Consequently, modelling makes sense only if there are order constraints between the event dates. In ChronoModel, a phase is defined as a group of target event dates, with no statistical model. The group of events may belong either to a 'stratigraphic' phase defined as a group of ordered contexts, or to a 'chronological' phase defined as a set of contexts built on the basis of, inter alia, archaeological, architectural, geological, environmental criteria. In practice, a 'context' is defined by the nature of the stratification at a site, and the excavation approach used by the archaeologist. Together, these two phases determine the smallest units of space and time (i.e. the context) that can be identified in the strati-graphic record at an archaeological site. Some target events are then dated in these contexts. We estimate the beginning, end and duration of a phase directly from the group of target dates, without adding any supplementary parametrisation. Within 'multi-phase' sites, each chronological or cultural phase can include distinct clusters of stratigra-phic units (Fig. 2) that could have been defined by spatial criteria, or in some cases, by the subdivision of a single phase deposits in distinct features directly (i.e. successive) or indirectly connected (e.g., covered by deposits from the following phase); in addition a single feature (e.g., stratigraphic unit) can provide one or several dates. For this paper, each site was modelled separately. As a result, the dates a posteriori for a multi-phase or mono-phase site are totally independent of those obtained from the others sites. Each phase modelling is based on three runs of Markov's chains of 10000 iterations each. In order to compare the modelled phases, we calculated the probabilities of anteriority for each phase's beginnings compared to the others. For each couple of phase beginnings [OA