Over the past 50 years, much of the work on obsidian in the West Mediterranean has consisted of analyzing particular archaeological assemblages to document how a specific analytical technique can be used to discriminate the products of the region's various obsidian sources (e.g., Bigazzi et al., 1986, Cann and Renfrew, 1964, Hallam et al., 1976 and Tykot, 1996). This tradition of scholarship continues today, being used to show that the various sources and subsources of the West Mediterranean can be distinguished with varying success using analytical techniques such as particle-induced X-ray emission (PIXE), neutron activation analysis (NAA), inductively coupled plasma mass spectrometry ICP-MS, Mössbauer spectroscopy, electron paramagnetic resonance, and XRF (Barca et al., 2007, Le Bourdonnec et al., 2005, Poupeau et al., 2010, Scorzelli et al., 2001, Seccaroni et al., 2008 and Tykot et al., 2013).

While this work represents a great success story in geoarchaeological research, questions of broader anthropological significance are less well developed, and it is only in the past two decades that more archaeologically driven questions have been forefronted (see Tykot and Ammerman, 1997). This is not to say that issues surrounding exchange mechanisms and the transportation of obsidian across the sea- and landscape have not been addressed, it is just that the results from so many analyses over the past 50 years make it almost impossible to determine the relative importance of different sources at any one particular site or how the uses of the various source products may or may not have changed through time (Pessina and Radi, 2006: 435). By focusing on one aspect of obsidian consumption, i.e. the raw materials involved, without considering the specific forms in which the obsidian circulated (nodule, core, end-product, etc.), the techniques by which they were worked, and the form of the final tools, the interpretive potential of much of this work is limited (cf. Freund, 2013).

While early obsidian sourcing studies in the West Mediterranean were limited to a small number of scholars, the past two decades have seen a proliferation of research on the subject. Many archaeologists of this new generation recognize the importance of combining sourcing data with other information about form and function of the artifacts themselves in order to address a wider range of questions about how obsidian was incorporated into the lives of the people who used it; this has taken form most notably in Sardinia, Corsica, Liguria, Calabria, and southern France (see Ammerman, 1985, Costa, 2007, Freund and Tykot, 2011, Léa, 2012, Le Bourdonnec et al., 2010 and Lugliè et al., 2008).

A central component of these analyses has been the reconstruction of source-specific obsidian reduction sequences and the specification of which parts of these sequences are represented at particular sites – i.e., whether an assemblage attests to on-site production commencing with the reduction of raw nodules or merely contains end-products. A central conclusion drawn from these analyses is that multiple knapping traditions existed across the West Mediterranean region over the course of the Neolithic, including pressure-flaked blade industries in southern France and Corsica (Bressy et al., 2008, Costa, 2006 and Léa, 2012), direct percussion blade technology in southern Italy (Diamond and Ammerman, 1985), and flake-based industries in Sardinia (Freund, 2014 and Lugliè et al., 2008). Nevertheless, these conclusions are based on a small number of total analyses and there is still much work to be done in terms of understanding how and where obsidian was acquired, reduced, and consequently used. This is why integrated characterization studies combining sourcing data with techno-typological analyses and usewear studies are so critical in interpreting the role of obsidian in larger socio-economic processes through time, and therefore a central component of this research.

Since the export of artifacts from Sicily was not permitted, all of our analyses took place on site at the Paolo Orsi Regional Archaeological Museum (Syracuse, Sicily) and the Luigi Bernabò Brea Regional Archaeological Museum on the island of Lipari during the summers of 2012 and 2013. Our study consisted of two parts, techno-typological classification and elemental analysis. Fig. 2 displays some examples of the analyzed artifacts.

A Bruker Tracer III-SD pXRF machine was used to analyze the artifacts. The artifacts were cleaned with water to remove any dirt or other contaminants that could affect the results of our elemental analysis. A filter was placed directly into the machine to enhance results for certain trace elements (Rb, Sr, Y, Zr, Nb) already shown to be successful for West Mediterranean obsidian sourcing (see De Francesco et al., 2012, Freund, 2014, Poupeau et al., 2000 and Tykot et al., 2013). The artifacts were placed on the top of the machine and analyzed for a period of two minutes. Source assignations were achieved by calibrating the raw analytical data against standard reference materials to determine the actual concentrations, the results then being compared with data generated from known geological samples using the same instrumentation. This can most successfully be illustrated through a graph of the element ratios of iron (Fe) and rubidium (Rb) to strontium (Sr; see Fig. 3). For those artifacts whose subsource designations were not immediately apparent, the calibrated concentrations of a range of elements were examined in further detail. Most importantly, we looked for the presence or absence of a high Sr peak that is characteristic of Canneto Dentro material.

In addition to artifact sourcing, each artifact was analyzed techno-typologically. This included recording the maximum length, width, and thickness on all of the artifacts as well as recording attributes pertaining to flaking type (platform, bulb, lip, etc.), to understand how the blanks had been knapped. Artifacts were also categorized as nodules, cores, flakes, blades, or angular waste, data that along with presence of cortex (divided into distinct percentage categories) allowed for the reconstruction of the obsidian reduction sequences. Only the percentage of dorsal cortex was recorded on the flake categories. This in turn allowed for the identification of the various forms in which obsidian entered the site prior to its reduction. Any form of deliberate modification in the form of retouch was also documented as a means of describing Stentinello tool types. Finally, raw material characteristics including color and the presence or absence of spherulites was recorded to see if certain materials were preferred at certain sites.

Because of time constraints and the large number of artifacts recovered from the sites of Rinicedda, Matrensa, Stentinello, and Megara Hyblaea, we decided not to elementally characterize and fully analyze the entire assemblages from these sites, but instead to select statistically representative samples for analysis. Because we conducted our work at two museums over the course of two years, we had to employ several sampling strategies depending on museum protocols and logistical constraints. Despite the varied sampling strategies, we have counts of the total number of artifacts from all sites as well as basic typological information about all of the artifacts, not just the sampled ones, upon which to make interpretations (see Table 2).

For the site of Rinicedda, all of the obsidian artifacts recovered from the excavation were bagged together according to the unit and stratum from which they were recovered; 20% of the obsidian artifacts were randomly selected for analysis. In order to get a representative sample of artifacts recovered from the site, all of the materials were laid out in their provenience-specific bags and 20% of the bags (i.e. every fifth) were randomly selected for analysis. Later, when we counted and categorized the remaining artifacts from the site, we analyzed any remaining cores. We analyzed all of the cores because they represent the first stage of reduction and were more likely to come from diverse subsources. For example, multiple flakes can be removed from the same core all having the same elemental signature; this is less likely to occur among cores.

For Matrensa and Stentinello, we began by analyzing all of the artifacts on display in the museum. Later, when we were allowed access to the rest of the collection from these sites, we divided the artifacts into categories based on their stages of reduction, including nodules, cores, flakes, blades, and angular waste. We then randomly selected a 20% sample of artifacts within each of the remaining categories, including flakes, blades, and angular waste. This was achieved by laying out the divided material on a table and selecting every fifth artifact.

Because we never had complete access to the entire assemblage from Megara Hyblaea at one time and we were told that there were a large number of artifacts, we decided to take a 10% random sample from each box we received. This was achieved by laying out the material from each box and then selecting every tenth artifact for analysis.

Table 1 displays the results of the elemental analyses. Of the 622 artifacts analyzed, 618 came from the Gabellotto Gorge subsource on the island of Lipari while four came from Canneto Dentro. The material from Canneto Dentro included an unretouched distal flake and an unretouched medial blade segment from Stentinello, an unretouched whole flake from Fontana di Pepe, and an irregularly shaped multi-platform flake core from Rinicedda.

While the elemental data show that the Gabellotto Gorge subsource was the primary outcrop exploited during this time, there is still a fair amount of variability within the subsource itself in terms of color and knapping quality. Fig. 4 displays the raw material characteristics of artifacts from sites in which 100 or more objects were analyzed. Although an even inter-site distribution exists between artifacts that contain spherulites and those that do not, there is more variation in terms of color. For example, at the site of Rinicedda there is a higher percentage of gray obsidian (69%) than black obsidian (31%). In mainland Sicily, it is only at Megara Hyblaea in which gray obsidian outnumbers black obsidian, making up 52% of the assemblage. This could suggest that the residents of Rinicedda were obtaining raw materials in a different manner than contemporary communities on mainland Sicily. This would make sense considering the site's relatively isolated location north of the source itself; however, more information about lithic reduction strategies is needed before more reliable interpretations can be made.

Recording the amount of cortex on an artifact can be informative when discussing the various stages at which raw materials enter a site (Andrefsky, 2009). If raw nodules were reduced on site, then one would expect to find lithic debris that includes both highly cortical primary reduction and later secondary and tertiary flakes with less cortex. Only the percentage of dorsal cortex was recorded on the flake and blade categories. Of the analyzed artifacts, 97% of them had less than 20% cortex. Sixteen artifacts had 20–80% cortex and only two artifacts had more than 80% (see Table 3). Although the only two artifacts that had more than 80% cortex came from the site of Rinicedda, 96% of the site's assemblage had less than 20% cortex. It is therefore clear that the majority of artifacts from all sites were in the secondary and tertiary stages of reduction and primary reduction occurred elsewhere, likely at the source area.

Artifact counts and basic typological data are summarized in Table 2. While artifacts from eight sites were analyzed, it is only at four sites that we see a true representation of Neolithic reduction sequences (see Fig. 5). This is because some of our analyses included artifacts on display at the museum that were biased representations of lithic reduction sequences at the various sites. We therefore focus here on assemblages in which we have robust enough samples to make meaningful interpretations.

Based on the results displayed in Fig. 5, it is immediately evident that blades and flakes are the primary artifact categories represented at the four sites. Of the 55 total analyzed cores, six were prismatic pressure flaked blade cores – found at all four sites (e.g., Fig. 2B), while the other 49 were irregularly shaped flake cores. A further analysis of Fig. 5 reveals that more cores and angular waste were found at the island site of Rinicedda than at contemporary sites on mainland Sicily. The high number of cores at Rinicedda (n = 39) is likely a reflection of the site's close proximity to Lipari, where residents had more ready access to raw materials from the nearby island. The high prevalence of angular waste could be a reflection of the modern excavation techniques being employed, which is why one must be careful about making interpretations based on artifacts recovered from early excavations that did not employ modern excavation techniques. Regardless, such limitations likely result in the loss of smaller reduction debris and therefore do not affect our interpretations.

The majority of blades from all sites were highly regular in thickness with parallel edges (Fig. 2D). The blade platforms were generally small and elliptical; lips on the ventral edge of the platform were common. For these reasons, we argue that blades from Stentinello period sites in eastern Sicily were produced using a technique of pressure flaking (see Andrefsky, 2005: 118–119; Pélegrin, 2012: 483). The high degree of regularity in shape and thickness in combination with the small elliptical platforms, the relatively diffuse bulbs of percussion, and the presence of lips on the ventral edge of the platform preclude direct percussion (see Andrefsky, 2005: 118–119; Crabtree, 1972: 44). They also lack the uniform curved profile and large platforms indicative of indirect percussion. This tradition of pressure flaking is mirrored in the reduction of chert from the same sites (e.g., Fig. 6A).

In a previous study of obsidian artifacts from southern Italy, Diamond and Ammerman (1985: 65) argue that the majority of blades from Stentinello period sites in Calabria were created using direct percussion with either a soft or organic hammer, ranging in width from 8–15 mm; these blades were characteristically wide, often reaching more than 20 mm in width. Despite possible differences in how these blades were produced, the artifacts analyzed in Sicily were similar. Based on our analysis of 286 blades from the eight Stentinello period sites, the average width was 14.4 mm with a standard deviation of 5.7 mm; 36 blades from five separate sites were more than 20 mm wide. The average length of whole blades (n = 128) was 34.7 mm with a standard deviation of 13.7 mm, indicating a large degree of variability.

Most of the artifacts were unmodified. Of the 286 analyzed blades, 20% were intentionally modified, mainly in the form of marginal retouch. Only 8 blades were invasively retouched, although none in the form of recognizable tool types. Of the 218 analyzed flakes, 17% of them displayed intentional retouch. Twelve of these flakes were heavily retouched and included denticulates, notched pieces, small perforators, abruptly retouched transverse scrapers, and a pièce esquillée (see Table 4; Fig. 6). Despite the low prevalence of retouch among flakes, they were still likely used for a variety of everyday tasks, especially considering that 89% of the cores analyzed were flake cores. It is also possible that a large portion of flakes were the result of blade core creation and maintenance. Further studies on obsidian usewear will help clarify how and where flakes were used.

Through a study of 622 obsidian artifacts from eight sites in eastern Sicily, it is shown that Lipari, specifically Gabellotto Gorge, was the predominant obsidian source exploited by prehistoric populations of the Early/Middle Neolithic Stentinello period (ca. 5600–4000 cal B.C.). Our results suggest that the primary reduction of Lipari obsidian occurred at the source area by local populations, where these raw materials were consequently brought from the island to eastern Sicily and Calabria in the form of preformed cores. In contrast to sites in southern Italy, communities in eastern Sicily maintained a widespread tradition of pressure flaked blade production. Despite similarities between Sicily and Calabria in the initial procurement of Lipari obsidian, we see no evidence to suggest that sites in eastern Sicily acted as such nodes of redistribution as appears to be the case in many regions of mainland Italy.

Although incomplete, the perspective that is emerging from the study of obsidian in southern Italy is particularly interesting and has great potential for exploring the formation of long distance exchange networks at the onset of the Neolithic. This study is therefore an initial step towards a more comprehensive interpretation of the nature of obsidian exploitation in Sicily and a more thorough understanding of how obsidian was distributed from the island sources of Lipari and Pantelleria. In the future, it will be necessary to acquire more data from well-dated contexts where modern excavation techniques have been employed in order to make more concrete interpretations about the differential roles various sites played in the exchange of raw materials across the island. Furthermore, we have demonstrated that the integration of data from all stages of an artifact's life history, including procurement, reduction, use, and discard, makes it possible to more fully appreciate the cultural and socio-economic conditions that mediated its use.