C’est au Néolithique ancien que la Sardaigne a été colonisée durablement, après peut-être quelques incursions au Paléolithique [1], [3], [4], [5] and [16]. On connaît actuellement quelque 74 sites datant de cette période [11]. Le site de Rio Saboccu est l’un des dix sites du Néolithique ancien récemment découverts sur une terrasse alluviale située au sud-ouest du massif volcanique du Monte Arci [9] (Fig. 1).

Bien que plusieurs îles de Méditerranée occidentale offrent de l’obsidienne (Fig. 2), celle des sites néolithiques de Sardaigne est d’origine exclusivement locale [7], [14] and [24]. Elle provient du Monte Arci, où elle affleure en position primaire, mais elle est également présente en position secondaire dans les plaines environnantes, sous forme de blocs et galets de dépôts fluviatiles [10] and [13]. Quatre types principaux d’obsidiennes ont été produits dans le Monte Arci, qui diffèrent par leurs propriétés physicochimiques [22], [23], [25] and [27] et souvent par leur aspect optique [11], [26] and [29]. De plus, elles présentent un cortex plus ou moins développé en source secondaire [11] and [14].

Le site de Rio Saboccu, implanté à proximité de sources secondaires exploitables d’obsidiennes, renferme plus de 1000 artefacts lithiques, essentiellement en obsidienne.

After rare and still debated human incursions during the Palaeolithic [1], [3], [4], [5] and [16], the island of Sardinia (western Mediterranean) was permanently populated only from the VIth millennium BC onward with the arrival of Early Neolithic (EN) men, first bringing the impressed cardial ware pottery. There are now up to 74 EN sites that are known all over the island [11]. Whilst evidence of the use of obsidian in pre-Neolithic contexts is still missing, like in the wider western Mediterranean region, it has been observed in a sizeable fraction of EN sites lithic industries, especially near to the Monte Arci volcanic complex. The few exceptions concern EN sites of northern Sardinia, where good lithic materials like flint offer an alternative to obsidian. During the last twenty years, ten open-air EN sites were discovered in the alluvial plains southwest of the Monte Arci massif, along Quaternary alluvial terraces of the Rio Mannu and Rio Mogoro rivers, with an average frequency of one per 4.5 km² [9] (Fig. 1). Their lithic implements are almost exclusively composed of obsidian.

Earlier studies have shown that the Sardinian ‘archaeological’ obsidians came exclusively from the Monte Arci, in spite of other natural occurrences in the Tyrrhenian island of Palmarola, the Eolian archipelago island of Lipari and in the southernmost Pantelleria, between Sicilia and Tunisia ([7], [13] and [24], among others; Fig. 2). However, provenance studies have to take into account four types of Monte Arci obsidians, differing both by their elemental compositions [25] and [27], their physical properties [22] and [23], and also partly by their visual characteristics [11], [26] and [29]. They have also to consider that obsidian was available to prehistoric men not only in ‘primary sources’ in their parent-rocks inside the Monte Arci or in their immediate proximity as dismantled materials, but also in ‘secondary sources’ in its surrounding plains as decimetric and pluri-decimetric cobbles [10] and [13]. Differences in macroscopic properties (size, shape, presence or not and characteristics of a cortex) of obsidian from primary to secondary sources may be used to refine provenance studies [11] and [14].

Rio Saboccu is one of the EN sites of the coastal plain southwest of Monte Arci, where obsidian of workable quality is still dispersed in nearby secondary sources. Its more than 1000 collected lithic artefacts are almost only made of obsidian, which offers a unique opportunity to investigate one procurement–reduction–use system elaborated by the first inhabitants of the island in the vicinity of raw material sources. After the Early Neolithic site of Su Carroppu [13], Rio Saboccu is the second Sardinian Neolithic assemblage whose industry is investigated with an integrated provenance/chaîne opératoire approach.

The Rio Saboccu site is located on a low elevation terrace dominating the San Giovanni Pond, at 4.5 km from the actual southern shoreline of the Gulf of Oristano (Fig. 1). This terrace, which extends from the foot of the Monte Arcuentu to the coastal pond, presents an average elevation of 6 m a.s.l., slowly decreasing northwestward. It ends abruptly by a 3-m-high scarp dominating the pond. This abrupt morphology results from the erosive action of the pond waters when blown toward the terrace by a strong regional wind (Mistral). It is in this way that the Rio Saboccu archaeological sites, situated on this scarp, were unveiled. This was not totally unexpected, as in its immediate vicinity, on some 10,000 m², modern ploughing dispersed on the terrace surface many lithic artefacts. The Rio Saboccu site is composed of two deep dug-out structures, S1 and S2, filled with a homogeneous anthropogenic dark-brown soil rich in organic remains, including charcoal fragments. They extend from the depth of about 0.3 m underneath the surface of the terrace to depths of respectively, 0.9 and 1.3 m. The apparent lateral expansion of the structures reaches no more than 9.5 m (S2). The uniformity of the filling suggests a fast formation process.

The S1/S2 structures contained a large number of lithic artefacts, but very few pottery fragments. The sherds present very simple shapes, mostly hemispheric bowls; their surfaces are not decorated. Anthracological analysis carried on charcoal from the bottom of the S2 structure profile indicates a partially wooded land with Buxus Sempervirens and Juniperus sp. The few animal remains are limited to terrestrial fauna, particularly goats. Two concordant radiocarbon dates of 6266 ± 48 BP (5341–5066 cal 2σ, AA-58899) and 6230 ± 60 BP (5320–5027 cal 2σ, Ly-3010) were respectively obtained from charcoals collected at the base of the S2 structure (calibration from the routine dataset intcal04.14c [21]). Therefore, they date the Rio Saboccu sites to phase III of the Sardinian EN, by the end of the VIth millennium BC.

At this stage of the excavations, is not possible to ascertain that the surface artefacts concentrations scattered upon the Rio Saboccu site area are related or not to its S1/S2 structures, in spite of great techno-typological affinities of the respective lithic assemblages and of their almost total absence of pottery. In any case, at the time of its EN occupation, the sea level was still lower than today and the Rio Saboccu campsite more extended in the direction of the pond. Thus, we can only observe now the remnant fraction of the terrace occupied by EN men.

The assemblage collected at Rio Saboccu S1 and S2 structures amounts to 1058 chipped-stone artefacts, among which 1047 are obsidians and the remainder are local cherts and chalcedony. The technological analysis of the entire collection clearly shows that obsidian reduction was almost entirely performed in situ, given the presence of all the main stages of the reduction sequence. The greatest part of the obsidian assemblage is debitage. A number of pieces showing portions of cortical surfaces bear evidences of the exploitation of obsidians both from primary and from sub-primary (close to the original geological context) deposits and from secondary alluvial deposits. A relatively high number of cores found in the site reveal that sourcing strategies were oriented in both cases towards little angular blocks or small rounded pebbles (maximum length of less than 80 mm). As revealed by the generalized plain type of butts with no or minimum preparation, reduction was carried out with the direct percussion technique with a soft stone and aimed mainly at producing flakes with a unipolar method and a recurrent shift of the striking platform. This strategy was usually performed through one or multiple 90° rotations of the piece. The 31 exhausted cores show a tendency to intensive and extreme raw material exploitation, until reaching the technical limits for further reduction. The laminar blanks (mainly bladelets) are rare and their general frequency keeps below twelve percent of the total lithic collection. Given the main character of expedient production that well describes the Rio Saboccu assemblage, very few formal tools can be recognized. Among them, the more standardized category is the one of geometric pieces, whilst other formal tools, mainly backed and double backed pieces, burins and scrapers, show a high variability, as they were produced from flake blanks of very different sizes and morphologies (Fig. 5).

For practical reasons, it is not possible to source by instrumental means all obsidians of large lithic series. By chance, in the western Mediterranean, the number of obsidian sources to consider is limited and it is relatively easy to identify the source-island of any archaeological obsidian to the naked eye (see, e.g., [26]). It could be determined in this way that at Rio Saboccu, as for other Sardinian Neolithic sites, all obsidians are of local origin. The distinction between the Monte Arci obsidian types, SA, SB1, SB2 and SC can also mostly be accomplished by visual observation [11], [13], [14], [26] and [28]. However, due to some facies convergences, mostly between the SA and SB2 types [28], it is necessary to call for a complementary instrumental way of characterization, typically from 15% to more than 30% of a site series [15] and [26]. At Rio Saboccu, 40 obsidians were of dubious origin from the point of view of their optical properties and, therefore, their provenance had to be determined by means of their elemental composition.

Thirty-eight artefacts [15] and 24 Monte Arci obsidians [14], [15] and [19] were analyzed by particle-induced X-ray emission (PIXE) at the AGLAE extracted-beam facility of the ‘Centre de recherche et de restauration des musées de France’ (Paris). Two other sets of respectively, sixteen [14] and seven (Table 1) geological samples were also analyzed by PIXE using the nuclear microprobe lines of a Van de Graaff accelerator and of the new AIFIRA facility of the ‘Centre d’études nucléaires de Bordeaux–Gradignan (CENBG)’. Twelve artefacts were measured in a non-destructive mode, the others and the geological samples from polished sections.

At AGLAE and AIFIRA, proton beams of 3 MeV and 1 nA were used to analyse an obsidian surface of about 0.5 mm². The fluorescence X-rays emitted by the samples were simultaneously collected by two Si(Li) detectors, respectively for the low-energy (Na, Al, Si, K, Ca,Ti, Mn, Fe) and high-energy (Mn, Fe, Zn, Ga, Rb, Sr, Zr) parts of the spectrum. As at the other nuclear microprobe of CENBG, only one detector was available; the ‘light’ and ‘heavy’ element contents were successively obtained from proton beams of 1.5 and 2.7 MeV, respectively [14]. To insure internal consistency, one Monte Arci obsidian (‘ARC-URS’) was analyzed in all runs. Data treatments were performed by the 2000 version of the GUPIX software [2] and [17]. It appears that only three Monte Arci obsidian types, SA, SB2 and SC, are present in the set analyzed by PIXE (Fig. 3).

The major element compositions (Na, Al, Si, K, Ca, Fe) of two other artefacts [15] and of 80 geological samples [8] were determined from polished sections by energy-dispersive X-ray spectrometry with a JEOL scanning electron microscope (SEM-EDS) operating at 20 kV at the ‘Centre de recherche de physique appliquée à l’archéologie’ (Bordeaux). Data treatment followed a Phi-Rho-Z approach. Binary plots of Al–Fe and Al–Ca contents show that these two artefacts are of the SB1 type (Fig. 4).

Finally, from visual and instrumental determinations, 56.2, 28.9 and 14.4% of the artefacts belong to the SA, SB2 and SC types, respectively. Apparently, SB1 obsidians, with only four artefacts (0.5%), were nearly not used at Rio Saboccu (Table 2).

As demonstrated elsewhere [6] and [15], the obsidian collection from Rio Saboccu S1–S2 structures can be considered a well-homogeneous complex as regards the technological features. The presence of conjoining/refitting pieces lying at different depths confirms the sedimentologic and pedologic impression of a brief formation time of the deposit. Thus, the collected assemblage allows us to get insights into the lithic sourcing strategies applied by the Rio Saboccu EN human group. From a general standpoint, all the lithotypes exploited at Rio Saboccu should be considered as local. The cherts and chalcedony can easily be found in the skeleton of the alluvial sediments around the site. They were formed during the Miocene post-Aquitanian diagenetic phase associated with the outcrop of underwater volcanic formations of the Masullas region, some 25 km eastwards from the site. The four obsidian types characterised at Rio Saboccu come from various deposits. Very rare pieces of SB1 type were occasionally gathered in their primary and sub-primary deposits, located in an inner area of the Monte Arci massif, at a minimum distance of about 17 km to the northeast. Among the other geochemical types, the most abundant SA obsidians were systematically acquired at their primary deposit of Conca’e Cannas, 16 km eastward, whilst the SB2 secondary and primary/subprimary deposits were closer, 9 km and 14 km northeastward, respectively. On the contrary, secondary deposits of SC obsidians occurred up to the site position and in its immediate vicinity and were (although but less intensively than the other types) the only ones exploited, as indicated by the incidence of well-rounded and evoluted cortical surfaces in SC pieces.

All these aspects bear evidence of a mostly non-opportunist exploitation of obsidian. In spite of the in situ availability of SC obsidians and of nearby SB2 obsidians, there is a marked preference for the more transparent and vitreous SA type, whose exploitation required however a major energy investment, given the greater distance of its source from the site. Apparently, it was not technical constraints that have guided the obsidian types’ selection, given the similar percentages of blade(let)s production and the homogeneous morphological–dimensional characters of the selected blocks and cobbles. To the contrary, the quasi-absence of SB1 obsidians, rich in phenocrysts and internal flaws, might be accounted for by its lower knapping qualities, as shown by present-day experimentation.

Cherts and chalcedonies, which exhibit quite variable knapping attitudes, are about as frequent as obsidian on the Rio Saboccu terrace. These raw materials, probably because they were less adapted to the EN men's needs, were however much less frequently used. We cannot exclude for them an opportunistic behaviour.

The raw material selection and the chipping reduction in practice at Rio Saboccu were shared by the other EN sites southwest of Monte Arci, as in the open-air sites of Pauli Putzu, Orri and Sa Punta (Fig. 1) [9]. SA, SB2 and SC obsidians are also well represented there, whilst SB1 obsidians are very rare to absent [11], and an intensive exploitation of secondary alluvial deposit is common, especially for the SC quality. The small size of the selected raw materials is a culturally-oriented choice, given the presence of a wide range of dimensions of pebbles and blocks both in primary and in secondary sources. The selection of small raw material units is in good agreement with the low percentage of blade production, which is a general tendency of the EN assemblages in Sardinia and in the wider Tyrrhenian region [11] and [20]. All the above-mentioned aspects and mainly the consistent percentage of SB2 among the small obsidian assemblages in the EN contexts of the western Mediterranean suggest the presence of structured social interactions between communities characterized by high levels of on-sea mobility, as supported by the close technological and decorative affinities in the pottery production [12] and [18]. Thus, in the composite distribution of Sardinian obsidians in the EN assemblages from the northern Tyrrhenian region, the human groups settled close to the Monte Arci source could have acted (although probably unwittingly) as a ‘filter’. Further studies coupling visual/instrumental sourcing with technological/typological determinations on obsidian artefact series from other EN sites of the western Mediterranean might allow us to test the reliability of this hypothesis.