Elemental Compositions from Iron Age and Roman glass in northwestern Portugal

The chemical composition that glass from primary workshops brings to the analysis of glass products from secondary workshops renders interpretation of place of manufacture of glass objects and trade connections difficult. This paper discusses the analytical results of a set of ninety-three archaeological glass samples obtained from nine Iron Age/Roman Period hillforts, as well as from materials from the Roman town of Bracara Augusta, and a Late Roman villa. The analyses conducted by the Elemental Analysis Facility at the Chicago Field Museum using Laser Ablation - Inductively Coupled Plasma-Mass Spectrometry, revealed the complexity of the problem with one main glass group but ten subgroups. All archaeological sites but one, are located in northwestern Portugal. The aim of this study is to add to the corpus of analytical evidence for northern Iberia and explore the interactions between hillforts in northwestern Portugal and possible recycling practices.

Keywords: Glass; Iron Age Hillforts; Roman; Analytical Groups; Social Interactions; Portugal.

Despite the ubiquitous appearance of glass fragments in Iron Age/Roman Period hillforts in Northwestern Iberia, research is still in its infancy (see Gomes 2012:90-92 and the chart on p. 88-90) [1]. Most Iberian ancient glass research has dealt with secondary workshops in major Roman urban areas or in the Iberian central and southern regions, often benefiting from burial contexts that may permit relative dating from associated objects, typological statements or status and gender insights, a situation unlike that of the Northwest where burial evidence of any kind rarely exists. The absence of osteological evidence is presumed to be due to the acidity of the granitic parent rock (e.g., Cosyns 2011; Cruz 2009, 2014; Petit-Domínguez et al. 2014; Pereira et al. 2015) [2-6]. The total number of hillforts (generally called castros) present in north ern Iberia varies widely, but there is general agreement that about 2000-2500 castros existed in Northwest Iberia at different points in time (Lourido 1993:83-84) [7]. A recent survey of Iberian hillforts north of the Douro River, resulted in a final dataset of 1279 Iron Age settlements (Bowers 2021) [8].This dataset was based on archaeological sites identified only as a “Fortified Settlement” or “Castro” on both Portugal’s Portal do Arqueólogo and the Patrimonio Galego datasets (2021:106). Archaeological site reports often mention glass finds, but although frequently quantified they are seldom described, rarely have associated relative or chronometric dates, and have not been chemically analyzed (Cruz 2009; Gomes 2012:90-92). Beads, partial or complete objects, and identifiable vessel forms are described (e.g., Alarcão and Alarcão 1963; Cardoso 1953; Gomes, 2012:81-85; Viana and Oliveira 1954: Ruano 1996) [9-12]. This is to say that we have a very poor notion of how much ancient glass has been found in the northwestern hillforts and even less of whether those artifacts could be Iron Age or Roman Period (Gomes: 2012:112-114, 125-126,135-136).

The present study results from the generous loan of glass samples from diverse archaeological sites in the northwest of Portugal. The respective site archaeologists chose the samples loaned, and many of the samples result from early excavations poorly documented, mixed stratigraphic layers and surface finds. The number of samples loaned per site also created a great disparity in the universe under analysis. In part, this reflects the site’s excavated area, but also the number of glass sherds found or documented. Despite those handicaps that make this overall collection of samples less than ideal, the study has the advantage of gathering samples from several hillforts as well as from different types of sites.

The study includes 93 samples of glass sherds and beads from several Iron Age/ Roman period hillforts as well as from the Roman town of Bracara Augusta and from one late Roman Villa (Fig 1). Dating ranges for the hillforts vary considerably, but most are dated between the 2^nd century BCE and the 5^th century CE. Bracara was established in the 1^st century BCE., as the capital of the Conventus Bracarangustanus and, for a while, it was the capital of the Roman province of Gallaecia (284-288). The Braga samples come from different localities all related to at least three workshops identified in Braga, and its relevance as the center of secondary glass production is undeniable (Cruz 2009a; 2014:58; Petit-Domínguez et al. 2013) [13-14]. Mario Cruz (2014) has reported on campanulate glass bowls from Braga and other Roman Gallaecian sites with comparanda found at several Iberian and French localities and dated between the 4^th to 6^th centuries CE, emphasizing local production and possible trade connections (e.g., Cruz 2009 v. 2: 232-237; 249). Aside from the local production of Black Glass objects, Cruz (2009: v. 2:280-282) lists some necklace glass beads and pendants likely manufactured locally. In other words, was raw glass imported into Northwest Iberia, and Portugal in particular, only to satisfy local needs during the Iron Age and Roman Period, or were local craftspeople involved in production for, albeit limited, export or intersite trade? The answer to the first part of this question is certainly that glass was imported to satisfy local needs. Cruz’s evidence from campanulate bowls indicates that at least some vessels were traded.

Samples

The glass samples generously provided by archaeologists and museums came from the hillforts of Alvarelhos (Trofa), 6; Bagunte (Vila do Conde), 11; Briteiros (Braga), 8; Cabeço (Boticas/Chavez), 2; Senhor dos Desamparados (Esposende), 1; São Lourenço (Esposende), 19; São Paio (Labruge, Vila do Conde), 5; Nossa Senhora da Paz (Marinhas/Esposende), 12; and Terroso, (Póvoa de Varzim), 2. Within this sample universe, Castro São Paio is the only coastal hillfort, while Castro do Cabeço is the only hillfort in the northeast of Portugal. Seventeen samples came from Braga, and 3 from the Late Roman villa at Agra do Relógio (São Paio de Antas). Most of these samples have broad stratigraphic contexts and few are associated with relative or radiometric dates.

Analytical technique

The analyses were carried out at the Elemental Analysis Facility (EAF) at the Field Museum in Chicago, USA, in 2013 (2 samples), 2015 (2 samples) and 2018-2019 (everything else). Prior to 2016, the EAF was equipped with a Varian Inductively Coupled Plasma - Mass Spectrometer (ICP-MS) that was replaced subsequently by a Thermo ICAP Q instrument. Both ICP-MS were connected to a New Wave UP213 laser for direct introduction of solid samples.

The parameters of the ICP-MS are optimized to ensure a stable signal with a maximum intensity over the full range of masses of the elements and to minimize oxides and double ionized species formation (XO+/X+ and X++/X+ < 1 to 2 %). For that purpose, the argon flows, the RF power, the torch position, the lenses, the mirror and the detector voltages are adjusted using an auto-optimization procedure.

For better sensitivity, helium is used as a gas carrier in the laser. The choice of the parameters of the laser ablation not only will have an effect on the sensitivity of the method and the reproducibility of the measurements but also on the damage to the sample. To be able to determine elements with concentrations in the range of ppm and below while leaving a trace on the surface of the sample invisible to the naked eye, we use the single point analysis mode with a maximum laser beam diameter of 100 micrometers, operating at 80 % of the laser energy (0.1 mJ) and at a pulse frequency of 20 Hz. A pre-ablation time of 20 s is set in order, first, to eliminate the transient part of the signal and, second, to be sure that a possible surface contamination or corrosion does not affect the results of the analysis. For each glass sample, the average of four measurements corrected from the blank is considered for the calculation of concentrations.

To improve reproducibility of measurements, the use of an internal standard is required to correct possible instrumental drifts or changes in the ablation efficiency. The element chosen as internal standard has to be present in relatively high concentrations, so its measurement is as accurate as possible. To obtain absolute concentrations for the analyzed elements, the concentration of the internal standard has to be known. The isotope Si29 was used for internal standardization. Concentrations for major elements, including silica, are calculated assuming that the sum of their concentrations in weight percent in glass is equal to 100 % (Gratuze, 1999; 2016) [15-16].

Fully quantitative analyses are possible by using external standards. To prevent matrix effects, the composition of standards has to be as close as possible to that of the samples. Different series of standards are used to measure major, minor and trace elements. NIST 610 is a soda-lime-silica glass doped with trace elements in the range of 500 ppm (SRM 610). Certified values are available for a very limited number of elements. Concentrations from Pearce et al. (1997) [17] are used for the other elements. Two other standards, with compositions mimicking ancient glasses, were manufactured by Corning: Corning glass B has the composition of a soda-lime glass made with mineral soda and Corning glass D has the composition of forest plant glass (Brill, 1999, vol. 2, p. 544) [18].

We obtained 102 compositions (see Table 1). In the rare case of polychrome glass samples, the different colors were measured individually. Some compositions were eliminated from the discussion due to different reasons. Certain compositions were typical of fairly recent recipes, and the samples were considered as modern intrusions irrelevant to this study focused on the Iron Age and Roman periods. Some glass samples were not glass. And finally, a few more samples had unusual glass compositions that did not fit into any known groups. First, we will describe briefly these compositions although they will be excluded from further discussion before giving more details about the different glass groups identified among the Iron Age/Roman period glass samples.

Modern compositions, non-glass compositions, unusual compositions

Sample GPO003 is a small dark blue facetted bead with a potash- lime (K2O-CaO) composition. It is colored by cobalt (454 ppm) with significantly high concentrations of nickel (360 ppm) and arsenic (3540 ppm). The style of the bead and its composition are typical of beads manufactured in Bohemia in recent periods with forest plant ashes. Similar beads were published in Burgess and Dussubieux (2007) [19]. They were found at a North American site dating from the end of the 18th to the 19th century.

Three samples, SP020, SP023 and SP022, are dark pieces of glass that have in common a high lime composition (CaO = 16 to 20 %). These glass fragments contain from 2.6 to 5.7 % of soda (Na2O), 4.1 to 4.8 % of alumina (Al₂O₃), 1.5 to 5.8 % of magnesia (MgO) and concentrations of potash that are 1 % or lower. In Coutinho (2016) [20], similar compositions were measured from containers found in Portugal dating from the 17^th and 18^th centuries, suggesting here also some intrusions of recent material.

Samples, GPO033, 042, SP008, 9, 21, 24, 27 and 37 are bright green fragments of glass with a soda-lime composition that have exceptionally high chromium concentrations, from 653 to 1101 ppm. For the other glass samples, the concentrations for this element range from a few ppm to a maximum of ~ 100 ppm. According to Freestone and Bimson (2003) [21], chromium is not a usual additive of ancient glass and its use, to produce the color green, was certainly very uncommon before the 19th c. CE. In addition to these exceptionally high chromium concentrations, some subtle differences in the composition of these samples confirm that they are recent glass. Despite the fact that these samples have relatively high lime concentrations (8.3 - 10.5 %), they have the lowest strontium concentrations of all the samples we analyzed (39 - 92 ppm). The chemical similarity of calcium and strontium facilitates the replacement of calcium by strontium in nature, which in glass incorporating lime from, by example plants or seashells, translates into an increase of the strontium concentrations with higher calcium concentrations. The use of modern materials obtained through chemical processes could explain this anomaly (Chopinet 2004) [22]. Sample SP019 does not have a high chromium concentration but it does have a high lime and a low strontium concentration (respectively CaO=12.0 % and 82 ppm). This sample could also be a more recent intrusion. We placed in the same category, sample TER01 with low strontium (45 ppm) despite having 10.5 % of CaO.

A strange looking and extremely thin colorless sample (GP006) has an unusual composition. It has only 28 % of silica, 25 % of alumina, 20 % of chlorine, 10 % of potash and 5 % of lime. Higher silica concentrations could have indicated a corroded glass but as concentrations for this constituent are rather low, it is possible that this material is not glass.

Samples GPO007 and GPO008 are silica-based material with high alumina and high potash concentrations. Their vitreous appearance suggested that these objects were a by-product of glass production or glass working but their compositions did not seem to support this hypothesis. We could explain the high alumina concentrations by some contamination from the clay of the crucible where the glass was melted, high potash concentrations could be due to the mixing of wood ashes with the glass but overall, the connection with the other compositions is difficult to understand and the nature of these samples remains mysterious. Two other samples (GPO009 and SP005), with a soda-lime composition, are different from the compositions that will be discussed below due to an exceptionally high magnesia concentration (7.4 % in GPO009) and a high concentration of Li (556 ppm in SP005).

General description of the main glass group and its subgroups

Once excluded all the glass compositions described above, 87 samples are left. Most of these glass samples have a fairly similar profile with high soda concentrations (13 to 21 %). Lime generally is the second most abundant oxide after soda (5 to 9 %). We are in presence of samples that are manufactured with soda-lime glass. Potash and magnesia concentrations are generally low (< 1.5 %) suggesting that a soda flux from a mineral origin such as natron was used (Shortland et al. 2006) [23] (Fig. 2).

Higher concentrations of potash and magnesia are usually associated with the use of soda plant ashes as a flux. The earliest glass, found in Mesopotamia and Egypt around the middle of the 2^nd millennium BCE, was manufactured using soda plant ashes. Starting around the 10^th century BCE, soda from mineral deposits (e.g. natron) replaced soda plant ash (Schlick-Nolte and Werthmann, 2003) [24]. Mineral soda glass was produced in Egypt and the Syro- Palestinian region and was distributed widely throughout the Mediterranean basin and beyond. The use of natron started declining around the 8th century CE and totally disappeared toward the end of 1st millennium CE (Phelps et al. 2016; Schibille et al. 2019; Tite et al. 2006) [25-27]. Recent research distinguishes several sub-groups for mineral soda-lime glass manufactured from natron during the first millennium CE. Looking at the concentrations of different oxides related to the source of silica: SiO₂, Al₂O₃ and TiO₂, it is possible to separate glasses made in Egypt from glasses made in the Levantine area at different periods (Freestone 2020) [28]. Figure 3 displays the calculated values of two ratios: Al₂O₃/SiO₂ vs. TiO₂/ Al₂O₃ and indicates the presence of 10 subgroups:
- High iron, manganese and titanium glass (20 samples)
- Egypt 1 glass (3 samples)
- Levantine 1 glass (6 samples)
- Roman Sb (3), Roman Mn (22 samples), Roman Sb-Mn (3 samples) and Roman no Sb-Mn glass (6 samples)
- High iron, manganese and titanium 2 (17 samples) and Foy 2.1 (3 samples)
- Low Alumina glass (4 samples)

The High Iron, Manganese and Titanium or HIMT glass

Twenty glass samples have a composition that matches the composition of the HIMT or High Iron, Manganese and Titanium glass group. It is one of the largest glass groups in this study. As indicated by its name, this glass group has compositions with significant quantities of iron (Fe₂O₃ = 1.2 % to 2.9 % among our samples), manganese (MnO = 1.5 to 2.3 %) and titanium (Ti = 1478 to 3121 ppm). In the case of iron, we excluded one high iron concentration (for sample GPO047), that was certainly due to a voluntarily addition to create a darker color rather than a natural presence of this element in the sand. This type of glass is dating mainly from the fourth to the fifth century CE (Foster and Jackson 2009; Foy et al. 2003; Mirti et al. 1993) [29-31] and it likely originates from northern Egypt (Schibille 2022) [32]. De Juan Ares et al. (2018) [33] report that this glass type was identified all around the Mediterranean basin, including the Iberian Peninsula, although it is quite scarce in the Syro-Palestinian region.

Two sub-groups were distinguished among the general HIMT glass group: HIMTa and HIMTb, an iron-rich variant (Freestone et al. 2018) [34]. From data obtained on material excavated in Spain, a threshold for the Fe₂O₃ to TiO₂ ratio of 5.4 was established for the separation of the HIMTa subgroup (< 5.4) from the HIMTb one (De Juan Ares et al. 2019) [35]. The HIMTa variant generally dates to the fourth and fifth centuries CE (Freestone et al. 2018) while the HIMTb sub-group appears slightly later at the beginning of the fifth century CE (De Juan Ares et al. 2019). Of the 19 samples placed in the HIMT glass group (after excluding GPO047), sixteen samples have a Fe₂O₃/TiO₂ ratio below 5.4 and therefore belong to the HIMTa sub-group, while three samples are above that threshold.

Egyptian 1 glass

Three samples, one fragment of glass adhering to a piece of ceramic (GPO028) and a bi-color bead with a dark body and yellow decorations (GPO048) have higher alumina and titanium concentrations. Samples GPO048 and GPO048Y are part of a black and yellow bead. GPO048Y contains 43.5 % of lead (measured as PbO) and 3.8 % of tin (measured as SnO₂). Glass sample GPO048 is black and contains 11.4 % of Fe₂O₃, 1.3 % of SnO₂ and 15.1 % of PbO. Lead and tin produce lead stannate that is a yellow opacifier. High quantities of iron are often present in black glass. The identical SnO₂/PbO ratios in both glass (0.86 and 0.88) suggests that the two oxides are certainly contamination of the opacifier from the yellow glass in the black glass. In order to obtain the composition of the base glass used for samples GPO048 and GPO048Y, it was re-calculated after excluding PbO, SnO₂ and Fe₂O₃ so it is possible to compare the composition of these two samples with data from the literature. Samples GPO028, GPO048 and GPO048Y belong to glass group Egypt 1 that was further divided into Egypt 1A, 1B and 1C.

Table 2:Reduced compositions of samples GPO028, 48 and 48Y compared to the averaged compositions of glass Egypt 1C (Schibille, 2022: 54-55).

We found more similarities between our samples and glass Egypt 1C (Table 2). Glass Egypt 1C is a glass that shows traces of recycling and is dated from the 8th century CE (Schibille 2022: 51, Table 3, 54-55).

Levantine 1 glass

Five samples belong to the Levantine I group (Freestone 2005) [36] that, as indicated by its name, was likely produced in the Syro- Palestinian region. The Levantine I glass seems fairly similar to the glass produced at Jalame that is dating from the 351-378 CE. (Brill 1988) [37] but it was also produced at Apollonia-Arsuf where this glass type represents Late Roman and Early Byzantine productions (5-7^th c. CE). Phelps et al. (2016) show that it is possible to distinguish the two production centers based on their Na₂O/SiO₂ and CaO/Al₂O₃ ratios (Fig. 4).

Those ratios vary widely for the Portuguese samples with 3 specimens falling in the area covered by the Apollonia production and 1 sample falling in the area characteristic of the Jalame glass. One sample is in the area where both productions overlap. With the exception of GPO044 (MnO = 0.03 %), all the samples have fairly significant concentrations of MnO (> 0.8 %) indicating voluntary addition. Brill (1988), Freestone et al. (2000) and Schibille (2022:84) show that the Jalame glass contains MnO concentrations above the natural level of this oxide while Apollonia glass have a systematically lower level for this oxide. This trend seems to indicate that the Portuguese glass samples are more likely to be similar to the Jalame glass than the Apollonia one with the exception of GPO044.

Roman glasses

Two groups with low titanium concentrations are separated based on the presence of either Mn or Sb; both are elements used to decolorize glass (neutralize the color induced by the natural presence of iron in the sand). Only three samples (SP038 and 43 and GPO021) fit the definition of Roman Sb glass as established by Schibille (2022:34) with concentrations of Sb₂O₃ higher than 0.25% and concentrations of MnO < 0.03 %. In these samples, antimony has concentrations ranging from 2689 to 2964 ppm while MnO is ~0.01 %. Roman Sb glass is dated from the 1^st to the 3^rd (Freestone 2021) [38] or 4th century CE. (Schibille, 2022). A recent publication, using hafnium isotopic data established the Egyptian origin of Roman Sb glass (Barfod et al. 2020) [39]. A much larger group of samples (21 objects) have lower antimony concentrations and higher MnO levels and fall into the Roman Mn glass group. This glass group is contemporaneous with the Roman Sb glass group (1^st - 4^th century CE) but instead of being an Egyptian production it has been recognized as a Levantine production.

Three samples contain both antimony and manganese in significant quantities. Schibille (2022:32) indicates that primary glass was either produced with addition of antimony or manganese, and that the concomitant presence of both elements in an object, results from the mixing or recycling of a glass containing antimony with a glass containing Mn.

HIMT2 and Foy 2.1

Another group of glass has concentrations of titanium that are lower than those of the HIMT group but higher than those of the Roman- Mn/Sb glass groups. They form two distinct groups (Table 3).

Table 3:Comparison of the HIMT2 and Foy 2.1 average compositions of the glass fragments from Portugal (P-HIMT2 and P-Foy 2.1) and the average compositions for these glass groups and glass group Foy 3.2 as found in Schibille (2022:40).

Samples GPO045, 46 and SP031 are part of the Foy 2.1 group that is assumed to have an Egyptian origin and that can be found at sites dating from the 5^th to 7^th c. CE. (Schibille, 2022: 40). The second group has an average composition that is similar to the HIMT2 glass samples more common in the 4^th c. CE. and considered as resulting from glass recycling (Foster and Jackson 2009; Schibille, 2022).

Low alumina glass

Four samples, all beads with a dark blue color produced by cobalt, have the lowest alumina concentrations (around 1.0 % or less) of all the glass samples considered (SPO014, SPO015, SP055 and GPO049) (Fig. 5). Low alumina natron glass is found during the Iron Age in Europe, at the end of the 1st millennium BCE (Van Strydonck et al. 2018) [40].

Trace elements vary in a wide range with SPO055 containing higher trace elements in general compared to the three other samples. When looking at some trace elements such as Ce, Zr, and Y that were found useful to determine the origin of the glass (Egypt or Levantine), a similar trace element pattern appears for the 4 samples when comparing the CeO/ZrO₂ and Y₂O₃/ZrO₂ ratios in the four beads. On Figure 6, we added samples with a known Egyptian origin (HIMT glass) and with a known Levantine origin (Levantine 1) to see whether the low alumina beads will match either of these provenances. The low alumina beads have similar ratios as the Egyptian HIMT glass.

This study yielded a large variety of glass groups which is consistent with the fact that we are looking at a wide range of contexts (Table 4). The dominant glass group, the Roman Mn glass, manufactured in the Levant, is dated from the 1^st to 4^th c. CE. The HIMT glass from Egypt and the recycled HIMT2 glasses are the two most important groups during the subsequent period (4th to 5th c. CE).

Table 4:Summary of the sample distribution among the different identified glass groups.

Table 5:Summary of the number of samples of each glass type per site.

Table 5 shows that three glass types, the Roman Mn glass (1^st- 4^th century CE), the HIMT glass (5^th-6^th century CE) and the HIMT2 glass (more abundant during the 4^th century CE) dominate the glass assemblage in this study. They are chronologically distinct, and it seems that one glass just replaced the previous one starting with the Roman Mn glass, then the HIMT2 glass and finally the HIMT. In some instances, secondary glasses would also be exchanged in parallel.

All the sites, except for Castro de Cabeço, are located within a circle with a 25kms radius (Fig. 7), therefore contemporaneous sites could have been part of the same trade networks and certainly would have had contact with each other due to their proximity. We note that all the sites with a significant quantity of studied samples yielded some early glass (Roman Mn). The same comment applies to the HMT2 glass. The HIMT glass is also quite common across the studied sites, but this type of glass is missing from Bagunte. Most of Bagunte’s radiocarbon dates place the site between the 2^nd century BCE and the early decades of the 1^st century CE., thus overlapping with other sites within the same radius, but Bagunte is a large site that has been minimally excavated. That likely accounts for the paucity of glass samples. The highest proportion of HIMT glass is found at Castro S. Lourenço/Esposente, maybe signaling a more intense activity at the later period. This site also yielded 2 samples of Foy 2.1 glass that is a later glass type (6-7^th century CE). Only three artifacts from Agra do Relógio were analyzed, and they all belong to later periods (one artifact is Levantine 1, 4^th-7^th century CE and two are Foy 2.1 artifacts). This accords to the information from archaeological materials from a late rural Roman housing complex (3^rd century CE) as well as from a later Suebi/Visigoth one (de Almeida 2013) [41].

Three of the glass samples from Braga were found in the context of a workshop (CTT) and are all HIMT2, which agree with the dating of the workshop (4-5^th century CE). Was this glass recycled or brought to Braga as raw material? This glass does not contain consistently high concentrations of trace elements such as Pb, Sn, Cu, Sb, etc., and this suggests the use of fresh raw material instead of broken objects.

Most of the samples analyzed are fragmentary and come from settlement sites with poor dating controls. The samples from Braga’s various locations also reflect salvage excavations within a historically layered and complex modern urban context. This situation is not unique (Ingemark 2014:235) [42] but it causes methodologic problems as well as problems of evaluation and perception of the quantity of glass materials present in the Roman colonial frontier of Northwest Portugal.

As diverse as this sample universe is, there are a few things we can infer about human behavior from the sample types and from their distribution as discarded objects. The majority of the glass beads in this sample set (8 out of 10) come from settlements. A perusal of Cruz Catalogue of Forms (2019: v.2) based on Braga’s excavated materials, shows that the few beads found were from production contexts, some used and some the result of manufacturing mishaps. Cruz, who seems to have found this noteworthy, states, “all the types of beads are exemplified by one single bead, from local production contexts” (277). Our sample size does not permit conclusions, but it does allow for acknowledgment of the difference.

The samples from hillfort sites are from single vessel fragments mostly found in dwelling structures and clearly not cached as special objects. These may be the remnants of mementoes or gifts given to, or acquired by, servants or by returning Romanized soldiers. We consider this a likely scenario because of the singularity of finds, generally a rim or body sherd, and their rarity. The number of glass fragments is minimal compared to that of sigillata vessels as well as the variety of forms of the latter. If this commerce of gifts bespeaks of status and even manifestations of power, then glass objects were either more valuable and less available or, alternatively, less desirable. We did not detect any clear geographical trends, but that could be a result of the sample’s geographic distribution and the sample size, which are both a result of what the sites’ archaeologists offered to, or could, submit. However, as stated above, all the sites, except one located farther east, are within a circle with a radius of 25kms, making it likely that they exchanged glass objects manufactured in Bracara’s workshops or elsewhere.

The samples from Braga clearly indicate the range of production capacity exemplified by different parts of vessels (see e.g., samples 0139, 0140, 0142, 0143, and 0145) and variety of local workshops (see e.g., samples 0142, 0143, and 0792) possibly representing some degree of specialization but certainly satisfying different customer’s needs. From Braga’s samples it is worth to highlight sample GPO0280 (Fig. 8) in which are embedded, in the hot worked glass, fragments of Athenian pottery (Gomes 2012:45, 53, 181). Granted, it can just be a result of the recycling of vessels, but it can also be inferred that Athenian vessel copies were being locally manufactured. Due to its higher alumina and titanium concentrations this sample was placed in the Egypt I subgroup, dated from the 8th century CE and indicative of glass recycling practices.

The three samples from the Late Roman rural villa of Agra do Relógio, show a similar discard pattern as that of the hillforts, in the sense that those are isolated finds, but likely resulting from vessels acquired for household use.

The number of compositional subgroups appears to indicate a complex market related to glass resource availability, possibly limiting workshop decision-making. In the social encounter of the client with the object, what mattered was that the object was made of glass, the novel, or not-so-novel, medium of the craft. Still, the number of HIMT glass samples also indicates that during the 4^th and 5^th centuries CE, the local market had fair access to likely Egyptian glass supplies, as did the rest of the Mediterranean basin. Bracara Augusta was being built during this period. Interestingly, during this same period, our study also indicates a preponderance of glass samples (Roman Mn group) from Levantine sources, and this twofold commodity supply pattern seems to extend to the 7^th- 8^th centuries CE.

Table 1:Elemental Composition Tables

We wish to express our appreciation to all the Portuguese archaeologists who gracefully contributed with their samples and patiently waited for results. Our special thanks to the D. Diogo de Sousa Museum staff for their samples and helpful information. This project was partially funded by NSF grant BCS#1321731.

No financial interest or conflict of interest exists for both authors.

1. Aarne (1964) Antti, and Stith Thompson, The Types of the Folktale: A Classification and Bibliography. Folklore Fellows Communications, No. 184 (Helsinki: 1964).
2. Abdul-Hakim, (1996) A Shawqi. Al-Amīrah Dhāt al-Himmah (Cairo: Al-hay'a Al-Misria Al-Aama lilkitab).
3. Anonymous (n.d.). Al-‘Amîr Hamzah al-Bahlawân known as Hmazah of the Arabs. (Arabic). 4 vols, (Cairo: n.d.; (Cairo: Maktabt al Gomhûriyyah, n. d..).
4. Anonymous, (n.d.) Qissat ez-Zîr Sâlim, Abû Lailah al Muhalahal al kabî (Cairo: Maktabt al-Gumhriyya, n.d.).
5. Anonymous (n.d.). Sirat al-malik Saif, 4 vols. (Cairo: n. d. Gomhouriyyah ).
6. Awad (197 ??), Louis, ez-Zîr?? In Arabic (Cairo: 197 ??). [Irretrievable records].
7. Burton (1894/1983), Richard F., ed., tr., Arabian Nights: The Book of the Thousand Nights and a Night (London, 1894/1983).
8. Garry, Jane. See El-Shamy 2005.
9. Caskel (1930), Werner. “Aijyâm al-`Arab; Studien zur altarabischen Epic.” In: Islamica. III. Fsc. 5. 1930. Suppl.
10. Gohary, M, Hasan El-Shamy, eds. trs. (1972). Qâmûs mustalahât al-’ethnolojyâ wa al-folklore (Tr. of A. Hultkranz’s General Ethnological Concepts). (Cairo: El-Marifff, 1972).
11. Haddad (2018) Abed Fahed, Farah Bsieso, Samer al Masri.
12. Hamilton (1981) Terrick, Antar: a Bedoueen Romance; (1981, First published 1819. Antar and ‘Abla: a bedouin: a bedouin romance). (Delmar, N.Y.: Scholars' Facsimiles & Reprints, 1981).
13. Heath (1984) Peter. “A Critical Review of Modern Scholarship on `Sīrat ʿAntar ibn Shaddad" and the Popular Sīra’”. In: Journal of Arabic Literature, (Leiden: Brill), 15: 19-44.
14. Hultkranz (1960) Ako. General ethnological Concepds. (Rosekilde and Bagger, Copehagen, 1960)
15. Ibrahim/Salim (1966) Nabīlah. Dirash muqaranah (A comparative study), 5^th (Cairo: al-Maktabah al-’Akadîmiyyad).
16. Ibrahim (1968) Nabī Sīrat al-Amīrah Dhāt al-Himmah : dirāsah muqārinah / taʼlīf Nabīlah Ibrāhīm.Published ،(1968] Cairo: : Dār al-Kātib al-ʻArabī, [1968].
17. Lane (1973), Edward William, An Account of the Manners and Customs of the Modern Egyptians (New York, 1973).
18. Le Gassick (1991) Trevor, ed., “The Traditional Structure of Sentiments in Mahfouz's Trilogy: A Behavioristic Text Analysis.” Reprinted in: Critical Perspectives on Naguib Mahfouz, Trevor Le Gassick, ed. (Washington, DC: Three Continents Press), pp. 51-70.
19. Littmann (1950) Enno, Ahmed il-Bedawi, ein Lied auf den Ägyptischen Nationalheiligen." In: Akademie der Wissenschaft und der Literatur: Geistes und Sozialwissenschaftlichen Klasse, Mainz (Wiesbaden: 1950, no. 3), pp. 50-123; also issued as an independent book, pp. 1-73; see pp. 61-62 and "Anmerkungen," nos. 6 and 8, pp. 122-123.
20. Lyons (1955) Malcolm Cameron. The Arabian epic: heroic and oral story-telling. 3 vols. (Cambridge: Cambridge University Press 1955).
21. Oliverius (1955) Jarslav. “Aufzeichunges uber den Basuu Krieg in der Kunstliteratur und deren Witeren twicklung in arabibischen Volksbuch über Zir Salim”. In: archiv orientani.  1985. [based on a copy by A. S. al-Habaal].
22. Oliverius (1975) Jarslav. “The Epic and Geniologicl Cyclization in Arabic Folk Book about Zîr Sâlim” IN; Acta Univertatis Carolinae-Philologica 5 (Prgue University: )1975.
23. Oliverius (1971) Jarslav. “Theme und Motiveim arabische Volkbuch von Zir Sâ” In: Archiv orietalini 31: 129-145.
24. Patai, Raphael (1969) Golden River to Golden Road: Society Culture and Change in the Middle East. (Philadelphia: 1969).
25. Sālem, Nabilah. See, Ibrahim, Nabilah.
26. Shamy (1976) Hasan, “The Traditional Structure of Sentiments in Mahfouz's Trilogy: A Behavioristic Text Analysis.” In: Al-CArabiyya: Journal of the American Association of Teachers of Arabic, 9: 53-74.
27. Shamy Hasan (1976) The Story of El-Sayyid Amad El-Badawî with Fatma Bint Berry, An Egyptian Folk Epic, part II, text and explanatory notes.” In: Folklore Forum, 11(3-4): 140-63.
28. Shamy Hasan (1976) The Story of El-Sayyid Amad El-Badawî with Fatma Bint Berry,” part I, “An Introduction.” In: Folklore Forum, (Indiana University), 10(1):1-13.
29. Shamy Hasan (1979) Brother and Sister. Type 872*: A Cognitive Behavioristic Text Analysis of a Middle Eastern Oikotype. Part I.1 of the “Brother-Sister Syndrome in Arab Culture”. (Folklore Monograph Series, Vol. 8, Folklore Publications Group, Bloomington, Indiana).
30. Shamy Hasan (1981) The brother-Sister Syndrome. In: International Journal of Sociology of the Family, Special Issue, The Family in the Middle East, Mark C Kennedy, ed., 11(2): 313-323.
31. Shamy Hasan (1995) Folk Traditions of the Arab World: A Guide to Motif Classification, 2 vols. (Bloomington, Indiana University Press, 1995).
32. Shamy Hasan (1999) Hasan, Tales Arab Women Tell: And the Behavioral Patterns they Portray. Collected, translated, edited, and interpreted by Hasan M. El-Shamy. (Bloomington: Indiana University Press, 1999)
33. Shamy Hasan (2004) Hasan, Types of the Folktale in the Arab World: A Demographically Oriented Tale-Type Index. (Bloomington: Indiana University Press, September 2004).
34. Shamy Hasan (2005) Hasan, “Sister and Brother (Motif, P253)”, pp. 349-61. In: Archetypes and Motifs in Folklore and Literature: A Handbook, Jane Garry and Hasan El-Shamy, eds. (M.E. Sharpe, 2005).
35. Shamy (2007) Hasan M, “Factors involved in Typological and Genre Studies”. In: Les Hommes et les recités: Théories et pratiques. Edited by A. Chrï Paris: L’Harmattan, pp. 315-361.
36. Shamy (2009): Religion among the Folk in Egypt. (Praeger: Connecticut, London: 2009).
37. Shamy Hasan (2007) Siblings in Alf laylah wa-laylah”. In: Ulrich Marzolph, Guest ed. The Arabian Nights: Past and Present. In: Marvels & Tales: Journal of Fairy-Tale Studies. Special Issue, Wayne University Press, Vol. 18: 2.
38. Shamy Hasan (2013) Beyond Oedipus: The Brother-Sister Syndrome as Depicted by Tale-Type 872*: A cognitive Behavioristic, demographically oriented, Text Analysis of An Arab Oikotype. (The Trickster Press, Bloomington: 2013).
39. Shamy Hasan (2016) Motific Constituents of Arab-Islamic Folk Traditions: A Cognitive Systemic Approach. 2 vols (Indiana University).
40. Shamy Hasan (2017/2016), Folklore of the Arab World”. Humanities: 2018, 7(3): 67.
41. Simpson (1972) William Kelly, ed., The Literature of Ancient Egypt. (New Haven: Yale University Press, 1972).
42. Taymûr (2000) Ahmad, Al-Amthâl al- Câmmiyyah (Vernacular Proverbs). (Cairo: Al-Shurûq, 2000)
43. Tayyib (1970) Al-Tayyib Mohammad, Al turâth al-shaCbi li qabeelat al-Humrân [Folk traditions of the Humrân tribe]. Mimeographed. (University of Khartoum, 1970).
44. Thompson (1964), See Aarne, above.
45. Tompson, (1955-58), Stith, Motif-Index of Folk Literature. (Blomington: Indiana University Press 1955-58).
46. Von Kramer (1803) Alfred, Aegypten, Leibzig, vol. I: 306-322.
47. Wannous (2017) Sadallah: Towards an Indigenous Arabic Epic Theater: An Applied Study of an Evening Entertainment and the Adventure of Slave Jaber's Head. In: (AWEJ for translation & Literary Studies vol,1(1).