Early Neolithic Pottery Production in the Maltese Islands: Initiating a Għar Dalam and Skorba Pottery Fabric Classification

One hundred and sixty-two sherds from the GD and SK facies, recovered from four archaeological sites on the islands of Malta and Gozo (Figure 1), were ground flat for stereomicroscopic analysis. Thirty-two of these sherds (18 GD, 14 SK, Tables 1 and 2), representative of the preliminary macroscopic and microscopic fabric groupings (described in Richard-Trémeau et al., 2023a), were selected for further petrographic analysis. When possible, three sherds characteristic of each group were selected. Petrography, supported by SEM-EDX and XRD, was carried out to identify fabric features (mineral composition and technological processes). These methods and the rationale behind using them are explored in sections below. All analyses provided material characterisation which could support, or negate, the hypothesis of local manufacture, and which provided indications on the composition of the paste and the firing temperatures attained. Descriptions and photographs of these sherds are available in repositories, along with photographs of the broader assemblages (Richard-Trémeau et al., 2023a, b, c).

In Maltese prehistoric studies, chrono-cultural phases and their associated pottery styles or facies are often named after the archaeological sites where they were first identified. For example, the term “Skorba” refers not only to the archaeological site (Figure 1, SKB or “Skorba site”) but also to a specific chrono-cultural phase and its characteristic pottery, which is now found throughout the Maltese Islands (SK or SK facies). This applies to most chrono-cultural phases of Maltese prehistory.

Diagnostic shapes in the GD facies include rims of Evans forms 3 and 4 (Figure 2a and b; Evans, 1971, figures 30–32), which are globular vessels with short tronco-conic necks; a body sherd with part of a strap handle (G1008) and other handling elements.

Figure 2

GD facies forms and examples of surface treatments in this assemblage. (a) Evans form 3; (b) Evans form 4; (c) G1002; (d) G1030; (e) G1004. SK facies forms and examples of surface treatments in this assemblage. (f) Evans RSk 4 (similar to GSk 2); (g) Evans GSk 4, RSk 6 has a similar globular body but with a straight rim; (h) Sagona GSk 11; (i) Evans GSk 3 (similar to RSk 5); (j) Evans RSk 7; (k) S1021 burnished; (l) S3002 slipped; (m) S6001 coarse SK, no surface treatment. Drawings not to scale, after Evans (1971) and Sagona (2015).

The SK facies vessels include Evans’ Red Skorba 4/Grey Skorba 2 forms (Figure 2f), which are fragments of pedestal bases; Evans’ Red Skorba 6 or 7 (Figure 2j) sherds, which have globular bodies with incurving or straight rims (Evans, 1971, Figures 31.2, 32.4.6–32.4.7). A possible ladle fragment (similar to Figure 2i) and several sherds bearing parts of handles such as knobs are also part of the assemblage (Table 2; see Sagona, 2015, pp. 38, 44 for illustrations of handle types in SK facies).

In cases where sherds lacked a diagnostic shape due to the fragmentation of the assemblages (Malone et al., 2020b, p. 323), features such as decoration were compared with existing literature (Evans, 1971; Malone et al., 2020b; Sagona, 2015; Trump, 2015). For coarse wares, which are typically undecorated, ware descriptions were used for validation (Evans, 1971; Sagona, 2015; Trump, 2015).

Eleven sherds feature decorations typical of the GD incised and impressed repertoire (Table 2; Cilia, 2004; Evans, 1971, p. 208; Malone et al., 2020b; Vella Gregory, 2021), including chevrons (e.g. G1002, Figure 2c), bands with repeated patterns such as curves or “C”s, and parallel lines (e.g. G1030, Figure 2d), as well as netting or grid patterns (e.g. G1004, Figure 2e). Three sherds show white-infilled decorations similar to Stentinello-style pottery (Figure 2d–e; Scarcella et al., 2011).

Three sherds from the SK facies have red surfaces produced by the application of a red slip (Table 2; Figure 2l; Evans, 1971, p. 210). The surfaces of two additional sherds (S3005, S3010; Table 2) also appear red, but macroscopically it is unclear whether this results from slip application or firing conditions. Red-slipped vessels form a minority within the wider assemblage (<20 out of >90 sherds) (Richard-Trémeau et al., 2023b, c); ten further examples have red surfaces and may also be slipped. Many vessels – slipped or not – show mottled surfaces. Burnishing is common in both facies.

The 32 sherds are sourced from four archaeological sites (Figure 1, Table 1, Table S1 supplementary material) located on the two main Maltese islands (Malta and Gozo). The pottery was obtained from the National Museum of Archaeology (NMA) collections in Malta. Descriptions of the 22 distinct archaeological contexts are available in the supplementary material (Table S2).

Seven GD and ten SK sherds were sampled from the site of Skorba in Mġarr, Malta (Figure 1 and Table 1). In the 1960s, two Late Neolithic megalithic structures were uncovered, and EN remains were further excavated, including Għar Dalam phase walls and Skorba phase hut-like structures (Evans, 1971, p. 19; Trump, 2015, p. 31, 1966). Recent excavations (FRAGSUS; Brogan et al., 2020a, p. 227) have aimed to refine the initial prehistoric chronological sequence (see section 2.6).

Eleven GD and two SK sherds were sampled from the site of Santa Verna, Gozo. Identified as a Late Neolithic megalithic structure in the early twentieth century (Ashby et al., 1913, p. 8; McLaughlin et al., 2020b), Trump later uncovered EN layers (as cited in Evans, 1971, p. 189). The most recent campaign aimed to reassess the site’s chronology (McLaughlin et al., 2020b, p. 123) and confirmed the presence of GD sherds alongside evidence of SK structures and floor levels with Late Neolithic disturbances (McLaughlin et al., 2020b, pp. 153–155).

The remaining two SK sherds were sampled from the sites of Taċ-Ċawla (Gozo) and Kordin III (Malta, Figure 1). Both sites feature Late Neolithic megalithic structures and were re-excavated recently. At Kordin III, explored in the early and mid-twentieth century, residual SK sherds were discovered mixed with later material (McLaughlin et al., 2020a, p. 221). Taċ-Ċawla was first explored in the 1990s (Malone et al., 2020c, p. 42). In the recent excavations, the EN presence at the site was primarily evidenced by scattered SK sherds (Malone et al., 2020c, p. 114). These two samples were included in the study to represent visible fabric variations within the SK sherds.

A fragment of each of the 32 selected sherds was embedded and processed to produce a thin section. These sections were then analysed using PLM. Fabric groups were established based on differences in the nature and frequency of inclusions and by examining the variability in textural and technological characteristics. Further subdivision of the main fabric groups into subgroups was based on variations in both inclusions and technological aspects.

Firing temperature was estimated by examining the degree of vitrification of the clay matrix and, in particular, the extent of dissociation of the calcitic inclusions. The type and grain size of the inclusions, however, would affect their dissociation, with micritic inclusions dissociating faster than larger crystals (Cuomo Di Caprio, 2017, pp. 232–233). It is expected that calcite dissociates between 800 and 900°C (Quinn, 2022, pp. 81, 435). The stability field of these inclusions may also be affected by the firing atmosphere and firing time (Maggetti et al., 2011, p. 506).

XRD, applied to powder samples of the same 32 sherds, was used to confirm previously identified phases, detect others beyond the capability of PLM, and compare sherd groups. XRD also aided in assessing firing conditions based on the presence or absence of mineral phases (Gliozzo, 2020; Quinn, 2022).

Fragments were ground into a uniform powder using a pestle and mortar (Garrison, 2014) and placed in a sample holder with the surface levelled and flattened. The sherd surfaces were not removed before grinding, nor was any component (e.g. calcium carbonate) extracted to enhance phase resolution. Intensity vs 2θ plots were smoothed using a simple moving average for noise reduction. High calcite content often obscured signals from other minerals, particularly silicates, except for quartz, making only the most prominent peaks visible. Additionally, the large number of potential mineral phases, many with complex lattices producing multiple peaks, and peak overlap further complicated identification. An example of this was the various sheet silicates, including kaolinite, smectite, illite, and glauconite.

Although reducing calcium carbonate content would have improved phase detection, the amount of material available for crushing was limited by practical and ethical considerations. The powder samples also needed to be preserved for future research and archiving. Parameters for the Bragg–Brentano configuration of a Bruker D8 are provided in the supplementary material (Table S5). Mineral phases were identified using Profex and the Rigaku PDXL software databases.

The sherd sections were imaged at low magnification (200× and less) and analysed using a Zeiss EVO MA15 VP-SEM with an Oxford Instruments X-MAX 50 mm² EDX. The aim was to correlate PLM observations with element detection.

Dates published by the FRAGSUS project suggest that evidence of Neolithic settlement in Malta could extend back to the early 6th millennium BCE (Table 3), although the earliest radiocarbon date directly associated with site activity and pottery is close to 5500 BCE (Hunt et al., 2020, p. 37; McLaughlin et al., 2020b, p. 146; 2020e, p. 35; Parkinson et al., 2021a, p. 212). Claims for an earlier farming occupation based on data from the Salina Deep Core have been rejected by Scerri et al. (2025, see 'Radiocarbon dating methods').

The dating campaign conducted by the FRAGSUS team was extensive in both spatial and chronological scope and questioned the previously established chronology for the EN (Table 3; McLaughlin et al., 2020e; Renfrew, 1972; Trump, 2002, p. 55). The new dating sequence pushed back both the phases of GD and SK earlier than previously established. The dating of the FRAGSUS team relies on evidence from the sites of Santa Verna and Skorba, which are the two main sites sampled by this study.

This article adopts the latest broad chronological boundaries for the EN (5500–4800 BCE) but still refers to the GD and SK as distinctive phases, for lack of a shared alternative explanatory model in the scholarship, while acknowledging that the latest evidence from the sites of Santa Verna and Skorba, reviewed below, tends not to support a strictly successive chronological phasing. The associated ceramics are termed “facies” to describe all the characteristics normally associated with each phase.

No pure GD phase layers were uncovered at Santa Verna and Skorba, so the FRAGSUS newly proposed dates of 5800–5400 BCE are primarily derived from the boundaries of the Skorba phase, which is assumed to follow GD. McLaughlin et al. (2020e, p. 31) further argue that these new dates align the GD phase with the broader spread of Neolithic Impressed Wares across the central Mediterranean. Recent modelling, including data from the central Mediterranean islands and mainland Italy (Scerri et al., 2025, Extended Data Figure 1), however, suggests that the onset of the Neolithic in Malta is later than in the rest of Italy and other Mediterranean islands. This latest modelling supports the boundary of the EN as mid-6th millennium BCE or after (7.4–7.1 ka). However, this model did not aim to resolve the issues surrounding the dating of traditional pottery facies since the data for each pottery facies are too limited.

The only published dates from a “pure” GD layer in the Maltese Islands come from Trump’s campaign at Skorba (1966), which produced dates of 5500–4700/5300–4100 cal. BCE (FB6, cited in Brogan et al., 2020a, p. 238; Table S4 in supplementary material). The GD pottery facies has not been dated since then, and Trump himself claimed that his dates, because of their limited number, were indicative at best (Trump, 1997, p. 174). Trump also claimed that his excavations produced pure GD layers overlain by pure Skorba phase layers.

The revised chronology for the Skorba phase (5400–4800 BCE; FRAGSUS) suggests that this phase began earlier than previously thought (Table 3; Parkinson et al., 2021a, b). However, all contexts in the FRAGSUS excavations had mixed ceramics, including residual GD pottery (McLaughlin et al., 2020e; Table S3). Some of the dated organic material may also be residual. These sites experienced extensive redeposition during the Late Neolithic occupation, complicating the chronological understanding of the earlier phases. Several contexts, from which the FRAGSUS dating material derives, belong to the first layers of activities overlying bedrock or sterile palaeosol. These mark the earliest datable occupation in the excavated trenches (Table S3).

There appears to be a hiatus in occupation at the sites (Skorba and Santa Verna), starting from 4800 BCE, with only one possible date of charcoal at the site of Skorba during the late 5th millennium BCE (Malone et al., 2020b, p. 312; McLaughlin et al., 2020e, p. 31). This hiatus suggests that attributing the early dates of the Skorba phase solely to residual GD material may not be a fully satisfactory explanation, as a mixing of activity from both phases would likely have resulted in a range of dates for both phases, and not strictly Skorba. Of note, Trump’s original chronology for the Skorba phase relied mostly on one date, which Trump himself considered “somewhat anomalous” (1997, p. 174) and overlaps with the Żebbuġ range (context LD5, Table S3). This means that the dating and boundaries of what are traditionally called the GD and Skorba phases are still not fully understood, and that the dates on which the previous chronology relied were not robust.

The most recent chronological model by FRAGSUS still differentiates the GD and Skorba phases as successive, although none of their excavations have corroborated this stratigraphic distinction, including at the site of Skorba (Brogan et al., 2020a). The assemblages claimed to be pure layers from the 1960s excavations may require reassessment and comparison with the assemblages found by the FRAGSUS team to determine if and where the form and stylistic boundaries of these facies are.

It is also important to note that these two pottery facies are not consistently found together in other contexts across the Maltese Islands (Richard-Trémeau et al., 2023d). For example, GD pottery has been found in cave contexts (Għar Dalam cave, Għajn Abdul, Scarcella, 2011), but the deposition of these sherds in these contexts remains undated. Therefore, the association of GD and Skorba ceramics is not systematic.

Considering the latest FRAGSUS excavations and chronology, questions can be raised about a possible contemporaneity or overlap between GD and Skorba pottery facies in Malta. The reluctance to explore this possibility stems from the need to situate Malta within the broader ceramic chronology of Sicily and Southern Italy. Other opposing arguments could include the pure successive layers in the excavations of the site of Skorba in the 1960s, the results of which have not been replicated, and the lack of systematic association between the two facies. Future research avenues could include: (1) reassessing and republishing the 1960s assemblages of these layers and the associated stratigraphic sections and sequences to better establish style and chronological boundaries of these facies; (2) redating material from Trump’s excavations directly associated with different pottery facies; (3) reassessing possible imports of Serra d’Altro and Diana identified by Trump at Skorba (mentioned by Evans, 1971, p. 211); (4) attempting direct ceramic dating, although the precision afforded by this might not allow a fine phasing resolution (Blain & Hall, 2016, p. 678).

The Maltese EN facies have often been compared to Neolithic traditions in Sicily and Calabria (Richard-Tremeau et al., 2023d). The GD facies has been linked to Stentinello pottery, primarily due to similarities in decorative repertoires, including incised and impressed motifs as well as white-infilled decoration. In turn, the Red Skorba facies has been compared in style to Diana ware, first identified at Lipari, but broadly distributed across Sicily (Dolfini, 2020; Parkinson et al., 2021b, p. 321). The stylistic resemblances between the Maltese and Sicilian wares (GD/Stentinello – SK/Diana) still need to be acknowledged, and the Sicilian wares offer comparative material, though they should not be equated (Vella Gregory, 2021). As Vella Gregory (2021) has argued, while the origins of some decorative techniques could derive from Sicilian traditions, the Maltese ceramic repertoire reflects local technological choices and communities of practice, which reworked and adapted these techniques in a distinct social and technological context. The Sicilian and Maltese wares may belong to similar technological traditions or spheres of knowledge exchange. However, the dynamics of these exchanges and influences cannot be fully assessed with the current evidence.

The chronology of the Sicilian and Calabrian Stentinello (and Stentinello-Kronio) phase, to which GD is often compared, is itself based on limited radiocarbon dates directly associated with the pottery facies (Giannitrapani, 2023, p. 159; Scarcella, 2011). The distribution of these dates is heterogeneous between western and eastern Sicily and Calabria (Tiné, 2014). Dates for the emergence of Stentinello-style pottery range from the early 6th millennium BCE at Uzzo (Scarcella, 2011, p. 42; Tiné, 2014), with the spread of the ceramic style beginning around 5700 BCE (Collina, 2015, p. 25). The dates also vary by specific region, for example, in Trapani, 5700–5200 BCE (Martínez Sánchez et al., 2016, p. 322), and in Lipari, 5500–5000 BCE (Martinelli et al., 2020). Stentinello-style pottery is described as having a long duration, coexisting with other pottery styles depending on the area. Several authors now include dates extending into the late fifth millennium BCE (Collina, 2015, pp. 25–26; Freund et al., 2015; Micheli, 2021) and suggest two phases (Quero et al., 2019; Speciale, 2024). Until the dating and ceramic evidence of Stentinello are synthesised to allow closer comparison between regions and overlapping facies (which is beyond the scope of this article), GD pottery can only be compared to Sicilian Stentinello ceramics through stylistic and technological analyses.

The FRAGSUS revised dating sequence also complicates comparisons between the SK and Diana-style wares. This resemblance was first observed at the site of Skorba, where it was determined that the pottery likely represents a local stylistic development – from Grey Skorba to Red Skorba – rather than imports of Diana ware, as previously believed (Evans, 1971, p. 209, revising his earlier claims). Recent dating of stratified contexts including Diana-style wares in Sicily suggests a range of 4300–3900 BCE (Parkinson et al., 2021a).⁽²⁾ More broadly, there is general agreement that Diana-style wares emerged in the mid-fifth millennium BCE or later (Giannitrapani, 2023, p. 161). However, the scarcity of recently dated stratified contexts from this period of the Sicilian Neolithic continues to limit the understanding of pottery chrono-cultural phases and contemporary stylistic developments (Giannitrapani, 2023).

PLM analysis, supported by SEM-EDX, identified three main fabric groups for each phase (GD and SK; Figures 3–7). Except for GD1 (Figure 4a), all groups indicate the use of temper (Table 4), as evidenced by well-sorted, coarse, and often angular inclusions (Cuomo Di Caprio, 2017, p. 103; Eramo, 2020). The identified tempers include spathic calcite (e.g. Figures 4c, h–i and 6g–i) and various types of crushed limestones (e.g. Figures 4f–g and 6a–c). Foraminifera were observed across the groups. Small, rounded glauconite pellets were detected, with a higher prevalence in the coarser groups (e.g. GD2, Figure 4e).

Figure 3

Photographs of the fabrics of the GD (left) and SK (right) facies as per PLM groups. Scale applied to all figures (the same magnification was used).

Figure 4

Microphotographs, XPL, for the GD facies fabrics. (a)–(c) GD1; (d)–(g) GD2; (h)–(i) GD3. (a) Abundant microfossils (both benthic and planktonic foraminifera belonging to rotaliids and globigerinids, respectively) and scarce calcite (GD1.1); (b) small benthic foraminifera (nodosariid) and few angular spathic calcite (GD1.2); (c) globigerinid planktonic foraminifera (cf. Trilobatus); (d) abundant fragments of biocalcarenite with rotaliids and nummulitid foraminifera; (e) spathic calcite, glauconite and fragments of calcarenite; (f) clast with a section of the foraminifera Operculina and glauconite; (g) biocalcarenite fragment and deformed foraminifera Operculina and globigerinid; (h) and (i) abundant angular spathic calcite.

Figure 5

(a) and (b) (GD1.1): abundant microfossils (ornatorotaliids and planktonic foraminifera Trilobus spp.) and minor calcite fragments; (c) (GD1.2): microfossils (Lenticulina sp. and planktonic foraminifera) and large angular calcite fragments; (d) (GD2.1): biocalcarenite and calcite fragments; (e) (GD2.4) abundant broken and deformed operculinid fragments; (f) (GD3): abundant spathic calcite and ornatorotaliid.

XRD analysis confirmed the presence of calcite and quartz, in all samples, though in varying proportions across and within groups (Table 5). Additionally, traces of minerals common in local geology – likely introduced with clay or temper – were identified in some sherds, including glauconite, and, in isolated cases, zircon, which can also be identified in thin sections.

Gypsum was tentatively detected in a few instances by XRD, but was absent in SEM-EDX analysis. As it is unlikely to survive firing, any presence likely results from post-depositional processes (Deer et al., 2013, p. 446; Quinn, 2022, p. 81). Possible hydroxyapatite was identified with EDX in group GD1.1 and may be linked to glauconite (Basso et al., 2008, p. 94).

Regarding firing temperature, all groups except Skorba SK1 and SK2 (Figures 6a–f and 7a–c) contain intact or partially intact carbonate components, including calcite and foraminifera tests. This indicates firing below the calcite dissociation threshold (T < 850°C) for the GD groups and SK3, while SK1 and SK2 were consistently fired at higher temperatures, closer to the threshold. No complete dissociation of limestone fragments or calcite monocrystals was observed. The peaks of clay minerals, such as illite, smectite, and kaolinite – present in Blue Clays (BC) (John et al., 2003; Pedley & Clarke, 2002) – were detected across all groups, suggesting relatively low firing temperatures and short firing times (Gliozzo, 2020; Quinn, 2022, p. 433). Wollastonite (>800°C) was tentatively detected in one GD and one SK sherd (G1006, S3004), whereas diopside (>800–900°C; Gliozzo, 2020; Quinn, 2022, p. 435) was absent (Table 5).

Figure 6

Microphotographs, XPL, for the Skorba facies fabrics. (a)–(c): SK1; (d)–(f): SK2, (g)–(i): SK3. (a)–(f): partial dissociation of microfossils (corallinacean red algae, small benthic foraminifera, nodosariids and rhodoliths) and limestone fragments; (b), (c) and (e), (f): red slip visible. (b), (c), (e), (f) have a red slip visible; (a)–(c): fragments of biomicrite, fossils and microfossils with incipient dissociation; (d)–(f): oolitic limestone with incipient dissociation; (g)–(i): angular spathic calcite and microfossils (including planktonic foraminifera cf Trilobus), glauconite in (h).

Figure 7

(a) and (c) (SK1): fragments of biomicrite and some foraminifera (ornatorotaliids); (c) (SK2): possible oncoids (clumps of coralline red algae); (d)–(f) (SK3): moderate to abundant calcitic temper with ornatorotaliid foraminifera as well as cibicidoids and globigerinids.

Other minerals detected occasionally included ankerite (S1003, S1021, S3004, and possibly G1005) and magnetite (S1021 and possibly S3008). The presence of magnetite could be related to reducing firing conditions (Gliozzo, 2020), while ankerite has been detected in local clays (John et al., 2003). Not all reduced sherds showed magnetite peaks, which could be due to the drowning of the signals by other minerals found in high proportions.⁽³⁾

The three GD fabric groups consist of a calcareous clay matrix and predominantly feature carbonate inclusions. Samples uniformly exhibit planktonic and benthic foraminifera, primarily globigerinoids and rotaliids, with occasional quartz and feldspars in the groundmass. Only one sample, G1005, displays partial dissociation of the carbonate components.

The 14 samples share a calcareous clay matrix with dominant calcareous inclusions, mainly microfossils (planktonic and benthic foraminifera) in varying sizes and quantities. Sand-size calcareous tempers differ in abundance, sorting, and coarseness, while silicate inclusions (quartz, feldspar, and glauconite pellets) are rare. These variations in temper define three fabric groups. The calcareous components of SK1 and SK2, including foraminifera tests, exhibit incipient destabilisation across all sherds. This indicates a temperature closer to the stability limit of calcite (about 850°C).

None of the raw materials observed across the different techniques indicated a provenance incompatible with the sedimentary geology of the Maltese Islands, although given the generic aspects of the sedimentary layers, it is only possible to state that there are no definitive imports in the assemblage. As in vessels from other periods (e.g. Anastasi et al., 2021, p. 3; Richard-Trémeau et al., 2024; Tanasi et al., 2020, p. 127), the matrices are highly calcareous, and foraminifera are common, particularly Globigerinoides (i.e. Trilobus spp.), Orbulina, and other planktonic foraminifera, common within the BC in Malta (Abels et al., 2005; John et al., 2003), as well as in other geological deposits. O. universa and O. suturalis (visible in SK3.1) date to the Middle Langhian, consistent with the upper layers of Globigerina Limestone (GL; Scerri 2019, p. 50) or the transition to the deposition of the BC layers (Prampolini et al., 2019, p. 119). More generic Neogene planktonic foraminifera were observed (e.g. SK 3.4), which is consistent with any geological layer in Malta’s post-Lower Coralline Limestone (LCL).

The primary temper materials identified in the analysed pottery include biocalcarenite and biomicrite, which align with the local Maltese limestones (Catanzariti & Gatt, 2014, p. 304; Scerri, 2019), and crushed crystalline calcite. Limestone temper (micrite) is present in ceramics from other periods (e.g. Tanasi et al., 2020, Bronze Age). These tempers likely originate from different geological formations across the Maltese Islands, with sedimentary rocks being intentionally crushed for use as tempers. Assuming that materials were collected from nearby locations to the sites where they were found, it is a reasonable hypothesis to assume the temper would have been obtained from Upper Coralline Limestone (UCL) (Table 12).

Figure 8

Geological maps with limestone members around the sites, as well as place names cited in the text. Sources include geological data from the Continental Shelf Department (Ministry for Transport and Infrastructure, Malta). Terra Rossa soils data are originally from Lang (1960) and were digitised by Alberti et al. (2018).

Calcite occurs naturally in small crystals within Maltese BC (John et al., 2003). As temper, large crystals may have been sourced from veins in the UCL and LCL formations (Reuther, 1984) or speleothems from caves (Pedley, 2005; Spiteri et al., 2010). Caves are known on the Xagħra plateau, where Santa Verna and other Neolithic remains are located (Figure 8) and close to Taċ-Ċawla. Sourcing crystalline calcite would have required knowledge of the locations of veins and/or speleothems and represents a specific raw material selection that extends across the two facies (GD and SK) and sites on the two islands.

Adding crushed calcite/carbonates as temper is a widespread practice during early phases of the Neolithic (from the seventh millennium BCE in southern Italy; Levi, 2010, pp. 190–192) and widespread examples across Italy (Capelli et al., 2008, 2017). Levi et al. (2019, p. 50) state that there was a decrease in the use of calcite as a temper after the Neolithic period in central/western Sicily. In other places, such as Lipari, local volcanic temper was added as it was the locally available material (Levi et al., 2019, p. 55) with greater experimentation during the Diana ware phase (Levi et al., 2020).

Albero Santacreu (2014b) suggests that using spathic calcite as a temper across the Mediterranean could be evidence of “contacts, social interactions and cultural transmission.” Calcite and limestone could have many properties, such as reducing the plasticity of the paste and improving workability (Albero Santacreu, 2010). Spathic calcite is also said to be thermally resistant and widely used for cooking wares (Eramo, 2020, p. 5; Müller, 2016, pp. 618–619). On the other hand, calcite is one of the few available tempers in Malta.

Taċ-Ċawla, in Rabat, Gozo, is on similar terrain to Santa Verna, about 1.5 km southwest, with comparable distances to geological sources (Figure 8).

SK2 samples are distinct due to their tempering with oolitic limestone. Oolites are found in the UCL (Pedley, 1978, p. 5: Ġebel Imbark Member; Scerri, 2019, p. 41: Għadira Beds part of Tal-Pitkal members; Figure 8), although a comprehensive review of possible locations is still needed. Ooids, both macroscale and microscale, have been sampled by Montalto (2010, p. 126) in BC from sites such as Għajn Tuffieħa (Malta) and Marsalforn (Gozo; Figure 8). Both oolitic sherds were found at the site of Skorba. While this temper may be related to pottery production close to Mġarr, this cannot be confirmed until additional EN sherds are sampled across the archipelago. No similar temper from this period (e.g. Pirani, 2018) or any period in the Maltese Islands has been identified in the literature. Oolitic temper is documented in the literature across Neolithic Europe (e.g. Jura, France, Martineau & Pétrequin, 2000; England, Peacock, 1969) and Bronze and Early Iron Age Europe (e.g. Black Sea region, Kulkova et al., 2023).

Looking at the material used for the decoration of sherds, Pirani (2018, p. 62) proposed that the red slip could be refined or purified clay with a similar composition to the matrix. However, another possibility that should be explored is the use of Terra Rossa soils, as these naturally red soils would contain fine angular quartz and calcareous inclusions in a fine red matrix (Richard-Trémeau et al., 2024, Figure 8). These are found, for example, across Mġarr above UCL and are also found as a sterile layer on the archaeological sites (e.g. Santa Verna: McLaughlin et al., 2020b, p. 146). Working with the raw material Terra Rossa would have involved new or different technological knowledge and would warrant further research.

Fine quartz is reported both in the local geology (French & Taylor, 2020) and observed in the bulk mineralogy. It is, however, only sporadically observed as visible inclusions in this study and other petrographic studies for the EN (Pirani, 2018, pp. 44, 47–48, 51), and in pottery fabrics from other periods (Bruno & Capelli, 1999, p. 61; Palmer et al., 2018, p. 582; Richard-Trémeau et al., 2024; Tanasi et al., 2015, p. 104).

Glauconite pellets are found in BC, possibly from the erosion of the Greensand layer (French & Taylor, 2020, p. 197). Glauconitic levels were described within some BC sections (Catanzariti & Gatt, 2014, pp. 309, 316, 320) and in Għajn Melel member sand pockets (UCL Bianco, 2017, p. 539; Pedley, 1978, p. 11). Glauconite grains are described in local fabrics (Roman pottery thin sections: Bruno & Capelli, 1999; SK phase thin sections: Malone et al., 2020a, pp. 743–749), as well as possibly added temper in Roman Imperial times (Bruno & Capelli, 1999). Different locations of clay procurement, either across sources or within a clay source (Richard-Trémeau et al., 2023d, p. 5), could explain different abundances of glauconite grains.

Mineral phases in the bulk mineralogy (XRD) may be attributed to local geological variability. The variations in the mineral composition of the Maltese sedimentary layers have not been widely explored, and barely any experimental works exist on the phase changes when these clays are fired at high temperature (for a student dissertation, see Xuereb, 2021). Comparisons with the local mineralogical literature are limited. These identified minerals could be present in the raw clay itself, from the erosion of Greensand and UCL layers situated above the clay, or from the added temper. Other identified geological minerals include zircon (found in GS, observed as inclusions with the SEM-EDX), hydroxyapatite in fossils or glauconite, dolomite (BC and GL, Gruszczyński et al., 2008, p. 239), and ankerite (John et al, 2003, p. 220). Other mineral phases, related to firing and mentioned in the results, include lime and wollastonite.

No archaeological evidence of early Neolithic ceramic production sites has been found in the Maltese Islands. Firing may have taken place in an open fire or pit fire, likely resulting in temperature and atmosphere variations due to stacking, fuel type, and weather conditions. Unlike other technological aspects, firing conditions did not remain consistent across the two EN facies.

All GD sherds, except G1005, were fired at a low temperature. This homogeneity in temperature might indicate a consistent firing regime. However, sherds in this assemblage have varying surface and fabric colours, suggesting a lack of uniformity in firing conditions. It is possible that keeping a low firing temperature could be intentional to prevent breakage caused by the change of phase of limestone and calcite (Eramo, 2020, p. 5).

This contrasts with fabrics SK1 and SK2, where higher firing temperatures are suggested by the incipient dissociation of the inclusions, a pattern that appears systematic for the two finer fabric groups. Most firing arrangements, including bonfires, are capable of reaching these temperatures, sometimes within minutes (Gosselain, 1992, p. 246). Achieving higher firing temperatures and properly firing the red-slipped vessel would have required specific knowledge and technical expertise (savoir-faire; Albero Santacreu, 2014a, pp. 56–57) to control the firing regime effectively and prevent breakage. The coarser and thicker-walled vessels of SK3 were, however, fired at a lower temperature, which shows a different technological approach, possibly because of different vessel functions.

In Lipari, Levi et al. suggest that Diana Ware involved experimentation, particularly with new methods of firing control to achieve red monochrome surfaces, as well as modifications in paste composition (Levi et al., 2020, p. 23). Robb further notes that Diana Ware in Sicily and southern Italy was fired at higher temperatures than earlier Neolithic pottery and was possibly fired in structures rather than open pits (Robb, 2007, p. 296).

The new fabric classification for the GD facies shows parallels with the classification of Trump (1966, p. 24, 2015), who distinguishes a fine ware (identified here as GD1) and a coarse ware (GD2). Trump also identified a transitional ware, which has the surface and form characteristics of the GD facies, but the fabric characteristics of the SK phase (“white grit”). This could overlap with the use of spathic calcite in GD3 as a temper (which looks like Trump’s “white grit”), and the use of other limestone temper in coarse wares in GD2. Trump described these fabrics based on the site of Skorba; however, in this current study, GD2 and GD3 are only found at Santa Verna. Alternatively, he might have been referring to all of the vessels tempered with white inclusions, which could overlap across all groups (apart from GD1.1) in this fabric classification.

Scarcella (2011) analysed 15 GD sherds from Skorba using PLM and other techniques. She identified two main petrographic groups based primarily on the abundance of inclusions. However, her classification grouped calcite crystals, micrite grains, and foraminifera (Scarcella, 2011, pp. 193–194), creating overlaps with the groups in the present study, which distinguishes fabrics based on the nature of these inclusions. Notably, Scarcella does not mention tempering for the GD sherds, instead suggesting the intentional use of naturally inclusion-rich clays. In contrast, she argues for deliberate tempering with quartz in Stentinello-phase sherds from Capo Alfiere, modern Calabria (see also Morter & Iceland, 2010).

For the SK fabrics, previous studies have consistently described a distinctive “white grit” (Evans, 1971, p. 209; Sagona, 2015, p. 35; Trump, 2015, p. 49). However, hypotheses suggesting crushed gypsum (Molitor, 1988, p. 240; Trump, 2015, p. 49) or chert tempering (Sagona, 2015, p. 35) are not supported by the findings of the present study. Gypsum tempering is unlikely, as it would decompose during firing (at about 300°C). Of note, this “white grit” is also present in the GD fabrics and is not a strict marker of SK facies fabrics.

The results largely align with observations made by Pirani in a Master’s dissertation (2018, p. 55). However, the results of this thesis remain unpublished, and the slides were not available for restudy. Pirani proposed that calcite and limestone were used as tempering materials. She analysed 23 sherds from the site of Skorba, particularly from the SK phase structures, using PLM among other characterisation techniques.

Pirani (2018, pp. 44–51) identified four fabric groups differing primarily in the nature and abundance of inclusions. She recognised fabric groups similar to those in this study, including a calcitic group comparable to SK3. She separated two groups containing limestone (similar to SK1), distinguishing between unslipped vessels with biomicrite and red-slipped vessels, suggesting that the latter might contain sparitic calcite inclusions. However, this distinction was not identified in the present study. Finally, Pirani observed one sherd with a greenish matrix, which she classified separately. This could be the result of overfiring, as similar effects have been observed in other periods (e.g. Roman vessels; Richard-Trémeau et al., 2024). The presence of oolitic limestone, however, was not noted by Pirani.