A petrographic and chemical analysis of Trinidad pre-colonial ceramics

ABSTRACT

This work presents an exploratory investigation into the production of pre-colonial ceramics found on Trinidad through petrography and chemical analysis with XRF and ICP-OES. Four main petrofabric groups are identified and described: a shell-tempered group, a sponge spicules group, a grog group and a micaschist/quartzite group. All evidence suggest an origin local to the island. Most of the petrofabric groups are consistent with ceramic series which were previously described, but never analysed petrographically and/or chemically.

GRAPHICAL ABSTRACT

KEYWORDS:

• Trinidad
• pre-contact
• ceramics
• petrography
• XRF
• ICP-OES

1. Introduction

Caribbean archaeology can look back on a rich research tradition. From the beginning of the discipline (mid to late nineteenth century) until roughly 1960, it was mostly concerned with describing, classifying and ordering artefacts, assemblages and sites to build large culture-historical typochronological sequences (Keegan and Hofman 2017, 17–20; Siegel 2013, 22). In this approach, the “culture-area concept”, which assumes that a culture is an adjustment to a particular environment, was the main tool to structure the archaeological data, along with “core–periphery relations” and migration waves (Siegel 2013, 23). Prime examples of this approach are J.W. Fewkes (Fewkes 1922) who defined twelve culture areas and three cultural epochs and M.R. Harrington (Harrington 1921) who postulated three waves of migration which populated the Caribbean. Perhaps the most influential figure in Caribbean Archaeology whose earlier work is also situated in this period, was Irving Rouse. His PhD thesis, Prehistory in Haiti: A Study in Method (Rouse 1939), proved a milestone for the field because it presented a way to classify and compare potentially related assemblages in a systematic way and because it introduced his method used to achieve this of modal analysis, which would for decades be frequently used, refined and debated both inside and outside of the field of Caribbean Archaeology (Siegel 2013, 24–25). Another milestone publication of this earlier period was that of F. Rainey (Rainey 1940) which focused on subsistence economics as a proxy to investigate cultural systematics and migrations, in his case of the so-called Crab and Shell cultures (Siegel 2013, 26–27).

As of 1961, a widening of research interests in the field of Caribbean Archaeology took place. Although many scholars, even to this day, still perform(ed) research belonging to the culture-historical paradigm, new areas of research such as time–space systematics, ethnohistory/ethnoarchaeology, symbolism/religion, archaeometric artefact analysis and provenance studies, physical anthropology, cultural heritage, environmental studiesand studies into socio-political organisation saw the light of day between 1961 and today and especially since ca. 2000 (Siegel 2013, 27–28).

Despite this rich research tradition, there is still a significant knowledge gap concerning the archaeometric investigation of ceramics from the region. Except the work presented in the special volume of 2008 in the Journal of Caribbean Archaeology, e.g. Conrad 2008; Crock 2008; Descantes, Speakman, and Glascock 2008; Fitzpatrick, Carstensen, and Marsaglia 2008; Hauser, Descantes, and Glascock 2008; Isendoorn, Hofman, and Booden 2008; Kelly 2008, very little archaeometric work on ceramics of the region has been published to date, with the notable exceptions of Pavia, Marsaglia, and Fitzpatrick 2013; Ting 2016; Lawrence, Marsaglia, and Fitzpatrick 2016; Scott 2018 and Ting 2018. Moreover, despite the fact that ceramics from Trinidad have been extensively studied in the past, e.g. Boomert 2000; Boomert, Faber-Morse, and Rouse 2013, none of the aforementioned archaeometric studies relate to ceramics from Trinidad.

Therefore, this article will construct an archaeometric baseline for Trinidad ceramics using a combination of thin section petrography and chemical analysis of both ceramics and clay samples from the island. This baseline will enable provenancing of Trinidad ceramics, from which further archaeological insights in terms of production and consumption patterns and related cultural processes can be construed. It will also be a useful addition to archaeometric data of the region, enabling further inter-island comparisons of pre-colonial Caribbean ceramics.

1.1. Geological and geographical setting

Trinidad is the southernmost island of the Lesser Antilles, with an area of approximately 4,828 km². It is situated just 12 km from South America (Venezuela), with which it was connected until the late Pleistocene/early Holocene (Boomert 2000, 17; Boomert 2014; Keegan, Hofman, and Ramos 2013, 5). Figure 1 shows the geographical relation between Trinidad, the South American mainland and the surrounding islands of the Lesser Antilles.

Figure 1. Situation of Trinidad in the wider region.

The Caribbean is a region which displays a high geological complexity with several plate boundary interactions, subduction zones, divergent plate boundaries and volcanism. It consists of four geological provinces with numerous subdivisions, of which the provinces of the Eastern Caribbean (Lesser Antilles and Barbados Ridge) and Northern South America (Venezuelan borderland) are the most relevant here.

The volcanic arc of the Lesser Antilles is a double arc system made up of a series of islands from Grenada in the south to the Anegada Passage between Anguilla and the Virgin Islands in the north. From Grenada up to Martinique, these two volcanic arcs are superimposed, whereas from Martinique northwards, the arc splits into an outer ridge with limestones (Limestone Caribbees) and an inner volcanic ridge. This inner ridge is a continuation of the volcanic Windward islands. The Windward islands and the inner ridge together form the Volcanic Caribbees (Draper, Jackson, and Donovan 1994, 8).

The Barbados Ridge is a forearc ridge which only emerges above sea level at the island of Barbados and is capped with limestones and deformed sedimentary rocks (Draper, Jackson, and Donovan 1994, 8).

The Venezuelan borderland is part of a mountain belt which marks the boundary of the Caribbean and the South American plates. The Venezuelan Antilles contain weakly metamorphosed volcanic and sedimentary rocks. This extends into parts of Trinidad (Draper, Jackson, and Donovan 1994, 9).

Trinidad itself is made up of five physiographic regions: the Northern Range, Northern Basin, Central Range, Southern Basin and Southern Range. The Northern Range consists of lower greenschist facies produced by the emplacement of large sheets of Mesozoic metavolcanics and metasediments into Paleogene flysch. It is subdivided into various formations consisting of limestones, (calcareous) phyllites, quartzites, slates, shales and sandstones. It also contains the only major outcrop of igneous rocks on Trinidad: the Sans Souci Formation. This formation consists of basalts and gabbros alongside sedimentary rocks (limestones, shales, sandstones and conglomerates). The other 4 regions can be grouped under the denominator of the “Trinidad Province”, which is a series of superimposed and deformed sedimentary basins with Cretateous carbonates and clastics, Paleogene flysch and Neogene marine and deltaic sediments. The series is disrupted by mud diapirs. Conglomerates, various types of clays, silts and sands, limestones, lignites, ironstone, arkosic quartzites, marls and mudstones, cherts and silicified wood can be encountered in the various formations of the Trinidad Province (Donovan 1994; Weber, Ferrill, and Roden-Tice 2001).

Figure 2 presents an overview of Trinidad’s surface geology, along with the location of the sites and clay sampling locations of this investigation.

Figure 2. Trinidad surface geology with discussed sites and clay sampling locations.

1.2. Chronological overview

Trinidad already shows signs of occupation from about 6000 BC (Ramos and Pagan-Jimenez 2013; Pagán-Jiménez 2015), with the pre-ceramic Ortoiroid series (Boomert 2016, 17). During the last centuries BC, a gradual arrival of the first ceramic “culture”, the Saladoid, is visible. It is unknown whether there was still an Ortoiroid presence on the island at this time, so the mechanism of ceramic introduction on the island is unclear. It is clear, however, that the spread of the Saladoid in the wider region was rapid and that they maintained strong inter-island interactions, resulting in a great uniformity in pottery despite some regional stylistic differences (Boomert 2000; Hofman and Reid 2014, 300–304). Important Saladoid sites on Trinidad are Palo Seco and Cedros, although Saladoid ceramics have been recovered from all across the island. Exchange with and gradual adoption of other ceramic traditions, such as the Barrancoid of the Lower Orinoco on the South American continent are also visible from the first centuries AD. These interactions are to be found especially on the island’s southwestern littoral at the sites of Erin and Quinam. At these sites, Barrancoid ceramics are found in close association with Late Palo Seco Saladoid ceramics, potentially suggesting that both wares were manufactured and used largely simultaneously. In terms of ceramic recipe Barrancoid pottery is frequently tempered with particles of deliberately crushed quartz. Erin ceramics, the local manifestation of the Barrancoid series, tend to be tempered with deliberately crushed quartz, fine quartz sand, sponge spicules, or a combination of both (Boomert 2000, 127–217).

In the period between ca. AD 500 and 700, great dynamism is perceived in the South Caribbean. In Bontour, shell-tempered ceramics with minimal decoration emerge which are quite distinct from the Saladoid-Barrancoid material, marking the transition to the Arauquinoid (Boomert 2000, 127–217; Boomert 2013, 141–149).

The final pre-contact ceramic series in Trinidad is the Mayoid, that was tempered with caraipé, a type of tree bark (Boomert 2010; Boomert 2013, 149–152).

Figure 3 shows a summary of the main ceramic pre-contact series of the island.

Figure 3. Summary pre-contact ceramic age chronological chart based on 14C dates. Adapted from Boomert 2000, 128.

1.3. Archaeological contexts

1.3.1. Cedros

Cedros is the type site of the Cedrosan subseries of the Saladoid and consists of a series of shell midden deposits which measure at least 0.85 ha in total. It was discovered by oil geologists in 1923 and afterwards excavated by Rouse in 1946. The deposits consisted of shells, Cedrosan Saladoid pottery, stone, bone and charcoal. In the top levels, some transitional ceramics to the Palo Seco complex were encountered. In addition, charcoal samples for radiocarbon dating were taken which yielded dates of 2140 +/– 70 BP (378–2 calBC¹) as the earliest and 1850 +/− 80 BP (18 calBC – 380 calAD) as the latest date (Boomert, Faber-Morse, and Rouse 2013, 19–25).

1.3.2. Palo Seco

Palo Seco is a multicomponent site and the type site of the Palo Seco complex, Cedrosan subseries, Saladoid series. A number of distinct shell middens were encountered. These middens showed a stratification of two layers, of which the bottom layer yielded pottery, stone, bone and shell artefacts exclusively of the Palo Seco complex whereas the upper stratum additionally contained some ceramics of the Erin (Barrancoid) and Bontour (Arauquinoid) complexes. A fragment of charcoal found in association with Cedros ceramics at the site yielded a radiocarbon date of 2060 +/− 80 BP (357 calBC – 119 calAD) and charcoal recovered in association with Palo Seco pottery yielded both slightly later dates of 1990 +/− 70 BP (194 calBC – 208 calAD) and 1480 +/− 70 BP (424–660 calAD) and a slightly earlier one of 2130 +/−80 BP (378 calBC – 6 calAD), although this last result may have been due to post-depositional factors (Boomert, Faber-Morse, and Rouse 2013, 25–30).

1.3.3. Quinam

The site of Quinam consists of three separate shell refuse deposits consisting of Palo Seco and Erin complex ceramics associated with similar material as on the previous sites. In another test pit at the same site but in a different midden, nearly 500 Bontour complex sherds were found (Boomert, Faber-Morse, and Rouse 2013, 30–35).

1.3.4. Erin

A multicomponent shell midden deposit measuring 260 m by 142 m was found in Erin which contained Erin and Palo Seco complex pottery next to stone and bone artefacts. In addition, a few Bontour complex pieces were found in the top levels of the midden (Boomert, Faber-Morse, and Rouse 2013, 40–42).

1.3.5. Bontour

Bontour is the type site of the Bontour complex and is also a shell midden deposit. Unlike the other middens discussed in this article, Bontour is not a multicomponent midden with multiple types of ceramics (Boomert, Faber-Morse, and Rouse 2013, 42–47).

1.3.6. St. Joseph

St. Joseph contains both pre-colonial and colonial remains. A 50 m by 50 m shell midden yielded pre-colonial Bontour complex ceramics as well as historic period Amerindian Mayo ceramics and European materials (Boomert, Faber-Morse, and Rouse 2013, 47–52).

1.3.7. Red House

As in St. Joseph, both pre-colonial and colonial ceramics have been found in Red House. The pre-colonial material found at the site is mostly Saladoid (Reid 2015).

2. Materials and methods

For this study, a combination of petrographic and chemical analysis was used. Thirty-three thin sections were examined, made from ceramics stemming from various locations on the island of Trinidad and all having a pre-contact date. Of these thin sections, one was made of a sherd from Bontour, one from St. Joseph, three from Cedros, three from Palo Seco, four from Erin, seven from Quinam and fourteen from Red House. All ceramics except the pieces from Red House were excavated by Rouse in 1946 or 1953 and curated in the Yale Peabody Museum until they were sampled by Boomert for archaeometric investigation. The material from Red House was donated by Reid from the collection of the University of the West Indies (Boomert, pers. comm.).

Different fabrics were isolated petrographically and afterwards representative thin sections for each fabric group were described in detail, using a descriptive method adapted from Whitbread (Whitbread 1995) and Quinn (Quinn 2013) and using reference charts for quantification of inclusion percentages. It was decided to keep the descriptions qualitative and to not use point counting because facilities needed for this technique were not readily available in the lab and the qualitative descriptions would later be complemented with quantitative chemical data, rendering much of the potential added value of point counting superfluous. Short comments about important differences between the described sample and the others within the petrographic group are included in the descriptions when applicable.

Normalised chemical data from powdered and homogenised samples obtained with XRF was provided by VU Amsterdam for the ceramic samples beginning with CB14 and the clays analysed, along with GPS-coordinates of the clay sampling locations. For the clays, this consisted of 22 samples whose locations are included in the map in section 1.1. In addition, ICP-OES analyses were executed at KU Leuven on the ceramic samples beginning with CB15 with a VARIAN 720-ES coupled with a VARIAN SPS 3 and ICP Expert II software. The operating conditions were 1.30 kW, 15 L/min plasma flow,1.5 L/min auxiliary flow and 0.75 L/min nebulizer flow, 3 s replication read time, a 60 s sample uptake delay and 15 s instrument stabilisation delay, a rinse time of 10 s and 3 replicates per measurement. PRI 1, GBW 7411, BCS 269, BCS 267, NIST 610 and NIST 612 were run as standards. The minimum correlation coefficient was set at 0.99.

XRF and later ICP-OES were selected as analytical techniques because they are destructive techniques which allow sample homogenisation and they give bulk chemistry results with the desired precision and accuracy (ppm range), see, e.g. Price and Burton 2012, 84–88; Isendoorn, Hofman, and Booden 2008; Vyncke 2011; Braekmans 2017. Moreover, they are in line with previous and future analyses within the research project and measuring facilities are available on-site.

The data obtained was then sorted according to the groups which were identified with thin-section analysis. The major element data was then first examined visually for striking inter- and intra- group differences or similarities. Afterwards, scatterplots were generated using Minitab 16 to assess this initial examination graphically. A similar procedure was used for the trace elements.

The online module of OxCal with calibration curve IntCal 13 was used for the calibration of the radiocarbon dates mentioned in the introduction.

3. Results

3.1. Petrographic analysis

Four main groups were identified, a shell inclusions group (N=3), a sponge spicules group (N=3), a grog group (N=21) and a micaschist/quartzite group (N=6). There does not seem to be a strong correlation between findspot and petrofabric group, except that the metamorphic group only contained samples from Red House and one sample from St. Joseph. In the following paragraphs, the different fabrics will be described in detail. Table 1 shows which samples are included in which group. The samples in bold italics are fully described.

Table 1. Overview of the thin sections.

Petrographic group	Petrofabric name	Sample code	Site
1	Shell inclusions	CB14BN483	Bontour
1	Shell inclusions	CB14BN466	Quinam, Excavation 1
1	Shell inclusions	CB14BN432	Cedros, Excavation 1
2	Sponge spicules	CB14BN459	Erin, Excavation 2
2	Sponge spicules	CB14BN452	Erin, Excavation 1
2	Sponge spicules	CB14BN480	Quinam, Excavation 4
3	Grog	CB14BN433	Cedros, Excavation 1
3	Grog	CB14BN442	Palo Seco, Excavation 1
3	Grog	CB14BN450	Erin, Excavation 1
3	Grog	CB14BN465	Quinam, Excavation 1
3	Grog	CB14BN470	Quinam, Excavation 2
3	Grog	CB14BN475	Quinam, Excavation 3
3	Grog	CB14BN478	Quinam, Excavation 4
3	Grog	CB15BN019	Red House
3	Grog	CB15BN022	Red House
3	Grog	CB14BN440	Palo Seco, Excavation 1
3	Grog	CB14BN445	Palo Seco, Excavation 2
3	Grog	CB15BN012	Red House
3	Grog	CB15BN018	Red House
3	Grog	CB14BN439	Cedros, Excavation 1
3	Grog	CB15BN014	Red House
3	Grog	CB15BN016	Red House
3	Grog	CB14BN456	Erin, Excavation 1
3	Grog	CB14BN476	Quinam, Excavation 3
3	Grog	CB15BN024	Red House
3	Grog	CB15BN025	Red House
3	Grog	CB15BN027	Red House
4	Micaschist/quartzite	CB14BN486	St. Joseph, Excavation 2
4	Micaschist/quartzite	CB15BN020	Red House
4	Micaschist/quartzite	CB15BN026	Red House
4	Micaschist/quartzite	CB15BN021	Red House
4	Micaschist/quartzite	CB15BN023	Red House
4	Micaschist/quartzite	CB15BN017	Red House

Note: The sections in bold italics are described in detail in this article.

Relative frequencies of inclusions and voids are determined using a percentage area reference chart and described using the terms shown in Table 2.

Table 2. Relative frequencies of inclusions and voids.

Predominant	>70%	Few	5–15%
Dominant	50–70%	Very few	2–5%
Frequent	30–50%	Rare	0.5–2%
Common	15–30%	Very rare	<0.5%

3.1.1. Shell inclusions group (N=3)

The sample from this group which is described in detail is CB14BN483 from Bontour. In terms of microstructure, the section contains very few (2–5%) voids of ca. 0.5 mm which are almost exclusively shrinkage voids around inclusions. The voids have a variable shape and roundness. The elongate inclusions and the elongate non-shrinkage voids which are present show an alignment to the edge of the sample. The matrix is optically active and has a pale to orange/brown colour. In terms of inclusions, the section contains 30% coarse fraction (>0.5 mm), 65% fine fraction (<0.5 mm) and 5% voids. The inclusions are angular and either very equant or very elongate. The size of the coarse fraction varies between 0.5 and 2.5 mm, with a mode of ca. 1.2 mm. The distribution of the inclusions is bimodal. The predominant, indeed almost exclusive, type of inclusion are shell fragments, conforming to the aforementioned characteristics. Apart from shell fragments, there are also very few (2–5%) argillaceous inclusions (clay pellets) present which are rounded and spherical. Their maximum size is ca. 1.2 mm and their modal size ca. 1 mm. The fine fraction consists almost exclusively of quartz and small calcareous fragments which are equant and angular.

In the other two samples from this fabric, shell is frequent (30–50%) rather than predominant (>70%). Other inclusions in these sections are grog (few) and limestone (few). CB14BN432 also contains very few (2–5%) plagioclase.

A photomicrograph under crossed polars of these three sections is included in Figure 4.

3.1.2. Sponge spicules group (N=3)

For this group, sample CB14BN459 from Erin is described in detail. Microstructurally, the sample contains voids between 0.2 and 2 mm in size which are mostly shrinkage voids of variable shape. There are so few inclusions that any alignment is impossible to detect. The matrix is optically active and light in colour. Moreover, it contains many very small fossils (smaller than 0.1 mm), most of which are completely circular, but some of which are elongate. These fossils are interpreted as sponge spicules. In terms of inclusions, the coarse:fine:voids (c:f:v:) ratio is 5:90:5, with fine fraction defined as inclusions smaller than 0.5 mm and matrix. The inclusions are sub-prismoidal or spherical and sub-rounded and the size of the coarse fraction ranges from 0.5 to 2 mm. The distribution is unimodal. In the coarse fraction, quartz is common (15–30%) and is elongate and rounded. The maximum size is ca. 1 mm and the modal size ca. 0.9 mm. Larger fossils and organic inclusions, the latter of which are visible from charring around some voids, are also common. Their shapes are variable and ill-defined and they have a maximum size of ca. 2 mm. Argillaceous inclusions (clay pellets and possibly mudstone or grog) are rare (0.5–2%) and also have an ill-defined shape and maximum size of ca. 2 mm. The fine fraction consists almost exclusively of quartz which is equant and angular.

The other samples from the group contain elongate rather than circular spicules and the matrix is orange rather than pale. In CB14BN480, a more or less parallel alignment to the edge of the sample can be detected. Otherwise, it is similar to CB14BN459. CB14BN452 displays only common (15–30%) spicules and common grog, so it is slightly ambiguous whether this sample should be included in this group or in the grog group. Figure 5 shows the sections under crossed polars.

Figure 4. From left to right and top to bottom: CB14BN483, CB14BN466 and CB14BN432 under XP.

Figure 5. From left to right and top to bottom: CB14BN59, CB14BN480 and CB14BN452 under XP.

3.1.3. Grog group (N=21)

This petrofabric group is internally heterogeneous. Sections CB15BN014 and CN15BN016 are described in detail as they represent the two main subgroups, of which subgroup 1 contains the majority of the samples (N=14) and subgroup 2 contains three samples: CB15BN016, CB15BN025 and CB15BN027. There are also three samples which deviate from the two main subgroups: CB15BN012, CB15BN024 and CB14BN439. CB15BN012 is similar to subgroup 2 but is also enriched in mica. The other two are similar to subgroup 1 but CB15BN024 contains large quartz inclusions and is enriched in mica and CB14BN439 contains lime-rich inclusions. Another interesting thing to note is that in many of the samples from subgroup 1, grog resembling the reducing part of subgroup 2 samples is present.

The microstructure of CB15BN014 contains voids of 0.5–1 mm, some of which are (sub-)discoidal and sub-angular while others are spherical and sub-angular. No alignment is visible. The matrix is light brown in PPL and yellowish orange in XP and is optically active. Some structures which resemble clay mixing – either intentional or natural – with a more grey/beige clay can be detected. However, these are probably soil textures rather than mixing (Neyt pers. comm.). The c:f:v ratio of the inclusions is 20:75:5 and the texture of the inclusions is sub-prismoidal or spherical and sub-rounded. The size of the coarse fraction ranges between 0.5–3.5 mm, with a modal size of ca. 0.7 mm. The distribution of the inclusions is bimodal and they are moderately well sorted. Grog is frequent to dominant (ca. 50%) and is equant and angular in shape. Its maximum size is ca. 3 mm and its modal size is ca. 0.7 mm. Other argillaceous fragments are few to common (ca.15%) and are equant and sub-angular. The maximum size is ca. 1.9 mm and the modal size ca. 1.2 mm. There are very few (2–5%) quartz inclusions present in the coarse fraction which are elongate and angular in shape and have a maximum size of ca. 1.2 mm and a modal size of ca. 1 mm. The inclusions in the fine fraction consist almost exclusively of quartz which is equant and angular with a size smaller than 0.5 mm.

The microstructure of CB15BN016 has voids which are ca. 0.05–0.6 mm in size and are either planar voids or vughs. No real alignment is visible, but the planar voids often occur along the rims of inclusions. Some optical activity of the matrix is visible in the oxidising region around the border of the sample, but the rest of the sample is optically inactive. The groundmass is grey with a yellow/orange border. In terms of inclusions, the section contains about 10% coarse fraction, 85% fine fraction and 5% voids, rendering the matrix very fine. The texture of the inclusions of the coarse fraction is sub-prismoidal or spherical and sub-rounded. The size of the inclusions of the coarse fraction ranges between ca. 0.5–1.8 mm, of which inclusions of ca. 0.7 mm are the most abundant. The coarse fraction is mostly situated in the oxidising border region. The distribution of the inclusions is bimodal and moderately well sorted. Grog and/or other argillaceous inclusions (clay pellets and mudstone) are frequent to dominant (ca. 50%) and the grog is equant and angular. The maximum size is ca. 1.8 mm and the modal size is ca. 0.7 mm. Distinguishing between grog and the other argillaceous inclusions is difficult. Quartz is frequent to dominant and comes in varying degrees of sphericity and roundness. The maximum size is ca. 1.2 mm and the modal size ca. 0.7 mm. The fine fraction is almost exclusively quartz conforming to earlier descriptions.

Pictures are included (Figure 6) of the two described samples and the three samples displaying variations in XP.

3.1.4. Micaschist/quartzite group (N=6)

The sample which is fully described for this group is CB14BN486. In terms of microstructure, it has voids of 0.5–3 mm, most of which are shrinkage voids of irregular shape. Other smaller voids are spherical and sub-angular. The alignment of voids and elongate inclusions is fairly parallel to the border of the sample. The matrix is optically active and orange/brown. In terms of inclusions, the c:f:v ratio is 20:70:10 and the inclusions of the coarse fraction are angular and elongate. The largest inclusions are ca. 3 mm and the mode of the coarse fraction is ca. 1.2 mm. The distribution is bimodal. Micaschist and quartzite are frequent to dominant (ca. 50%) and elongate and angular. The sample contains both inclusion types in more or less equal quantities. Their maximum and modal size corresponds to the general coarse fraction description. In the coarse fraction, there are also rare to few (<10%) argillaceous inclusions (clay pellets) present which tend to be elongate with varying roundedness. Their maximum size is ca. 2 mm and their modal size ca. 1 mm. The fine fraction is mostly quartzite which is equant and angular with a size smaller than 0.5 mm.

In CB15BN021 and CB15BN023, quartzite is more abundant than micaschist. CB15BN017 also contains abundant quartz in addition to quartzite and micaschist, but the fine fraction is more pronounced with a c:f:v ratio of 10:80:10. Figure 7 presents an overview of all six sections in XP and low magnification. Also note that the more or less parallel alignment is not visible in the other sections. Indeed, CB15BN020 displays a coil-like pattern instead.

Figure 6. From left to right and top to bottom: CB15BN014, CB15BN016, CB15BN012, CB15BN024 and CB14BN439 in XP.

3.2. Chemical analysis

3.2.1. Major elements

Table 3 shows the major oxide data for sampled clay sources and ceramics. The most obvious characteristic is that the CaO values of the shell tempered group are much higher than for the other fabrics or the clays, compensated for by much lower SiO₂-concentrations (closed sum). The samples of group 2 are also chemically consistent apart from sample CB14BN452. As noted before, the assignment of the latter to either group 2 or 3 is ambiguous and this is reflected in its chemical composition, which is far more comparable to group 3 than to group 2 (e.g. lower amounts of Na₂O and Fe₂O₃), despite the presence of some sponge spicules. Apart from some individual elevated or low values (e.g. P₂O₅ values between 4–6 wt%), groups 3 and 4 appear to be similar, albeit internally heterogeneous. Since multiple measurements from the same clay deposit are not available, it is unknown whether this heterogeneity was already present from the source, or is due to clay treatment such as mixing.

Table 3. Major oxides of ceramics and clays.

Sample n°	Group	Location	Na2O	MgO	Al2O3	SiO2	P2O5	K2O	CaO	TiO2	MnO	Fe2O3	Total
TRI-06		TRI-06	1.06	0.31	19.24	69.02	0.05	2.65	0.01	1.12	0.01	6.52	100
TRI-10		TRI-10	1.09	2.38	19.85	59.50	0.43	2.44	4.61	0.91	0.07	8.71	100
TRI-11		TRI-11	1.21	2.62	17.67	66.18	0.18	2.95	0.75	1.03	0.12	7.26	100
TRI-12A		TRI-12A	1.09	1.95	15.32	73.98	0.11	2.36	0.71	0.90	0.02	3.54	100
TRI-12B		TRI-12B	1.08	2.29	17.04	70.92	0.10	2.66	0.68	1.08	0.02	4.13	100
TRI-12C		TRI-12C	1.06	2.39	17.10	70.96	0.09	2.54	0.57	1.06	0.02	4.18	100
TRI-13		TRI-13	0.33	1.17	17.92	73.35	0.24	1.96	0.48	1.07	0.06	3.36	100
TRI-14		TRI-14	0.77	1.77	19.69	65.44	0.11	2.49	0.25	0.97	0.02	8.45	100
TRI-15		TRI-15	0.54	2.48	21.20	63.81	0.08	2.99	0.57	0.97	0.06	7.27	100
TRI-17		TRI-17	0.55	2.65	20.22	62.43	0.15	3.14	0.42	0.95	0.07	9.36	100
TRI-18		TRI-18	0.35	2.06	22.24	60.60	0.17	2.52	1.17	0.85	0.14	9.85	100
TRI-19		TRI-19	0.38	1.01	19.96	61.99	0.11	2.09	0.60	1.03	0.11	12.69	100
TRI-20		TRI-20	0.38	1.05	20.50	65.38	0.11	2.32	0.44	1.12	0.07	8.59	100
TRI-21		TRI-21	0.62	1.27	15.49	73.56	0.11	2.05	0.43	0.79	0.07	5.61	100
TRI-22		TRI-22	0.44	2.00	19.47	68.11	0.07	2.50	0.55	1.01	0.06	5.77	100
Average			0.73	1.83	18.86	67.02	0.14	2.51	0.82	0.99	0.06	7.02
Standard Deviation			0.33	0.71	2.03	4.84	0.10	0.34	1.08	0.10	0.04	2.66
Sample n°	Group	Location	Na2O	MgO	Al2O3	SiO2	P2O5	K2O	CaO	TiO2	MnO	Fe2O3	Total
CB14BN483	1	Btr	0.52	1.51	12.75	39.21	0.84	1.95	35.67	0.83	0.02	6.70	100
CB14BN466	1	Qui	0.57	1.13	17.38	47.46	0.99	1.49	22.36	0.84	0.04	7.74	100
CB14BN432	1	Cedr	0.33	1.24	17.83	57.97	0.50	2.01	10.35	1.03	0.08	8.66	100
CB14BN459	2	Erin	0.06	0.69	18.03	72.97	1.66	1.41	0.60	0.98	0.11	3.49	100
CB14BN452	2	Erin	0.32	0.65	21.27	61.66	1.25	1.80	1.88	1.12	0.06	10.00	100
CB14BN480	2	Qui	0.13	0.55	16.38	72.49	1.75	1.67	0.45	0.98	0.16	5.42	100
CB14BN433	3	Cedr	0.30	0.91	20.69	61.83	0.80	2.33	2.99	1.20	0.08	8.89	100
CB14BN442	3	P.S.	0.40	1.40	21.18	59.43	0.91	2.29	2.45	1.01	0.02	10.90	100
CB14BN450	3	Erin	0.72	1.48	19.07	62.00	0.72	2.09	2.20	1.03	0.12	10.57	100
CB14BN465	3	Qui	0.37	1.37	19.66	62.32	1.30	2.49	2.29	1.01	0.08	9.11	100
CB14BN470	3	Qui	0.32	1.14	20.49	61.52	1.24	2.06	1.92	1.10	0.08	10.12	100
CB14BN475	3	Qui	0.51	1.62	20.78	60.43	0.22	2.72	2.77	1.04	0.24	9.68	100
CB14BN478	3	Qui	0.53	1.80	21.10	60.44	0.80	2.59	5.78	1.14	0.03	5.78	100
CB15BN019	3	R.H.	0.46	0.62	25.77	52.67	6.08	2.46	3.21	1.06	0.02	7.63	100
CB15BN022	3	R.H.	0.58	0.75	22.42	59.78	3.47	2.24	2.66	1.06	0.04	7.01	100
CB14BN440	3	P.S.	0.84	1.61	18.71	65.59	1.27	3.19	1.92	1.06	0.05	5.76	100
CB14BN445	3	P.S.	0.28	1.34	21.46	66.10	0.40	2.39	1.35	1.17	0.04	5.45	100
CB15BN012	3	R.H.	0.63	1.01	21.68	59.02	3.43	2.07	2.66	1.04	0.03	8.43	100
CB15BN018	3	R.H.	0.50	0.66	25.90	54.18	6.04	2.21	3.70	1.00	0.02	5.78	100
CB14BN439	3	Cedr	0.22	1.21	20.40	60.99	0.87	2.01	2.77	1.04	0.20	10.29	100
CB15BN014	3	R.H.	0.46	0.72	21.91	59.95	4.13	1.96	2.42	0.98	0.04	7.42	100
CB15BN016	3	R.H.	0.38	1.08	24.74	56.60	4.03	1.78	1.59	1.16	0.05	8.60	100
CB14BN456	3	Erin	0.54	1.04	20.07	65.47	0.70	1.90	1.91	1.09	0.06	7.22	100
CB14BN476	3	Qui	0.67	1.46	19.33	65.52	0.28	2.23	1.35	1.05	0.04	8.08	100
CB15BN024	3	R.H.	0.47	0.54	17.41	66.87	3.25	1.79	2.35	0.86	0.02	6.45	100
CB15BN025	3	R.H.	0.42	0.96	22.97	57.44	3.10	2.18	1.99	0.95	0.04	9.96	100
CB15BN027	3	R.H.	0.43	0.74	24.83	55.84	4.47	1.83	2.25	1.13	0.04	8.43	100
CB14BN486	4	St. J	0.51	1.03	16.06	67.59	2.61	2.25	1.57	0.73	0.03	7.61	100
CB15BN020	4	R.H.	0.41	0.79	21.51	63.77	1.32	2.66	2.19	0.87	0.03	6.44	100
CB15BN026	4	R.H.	0.43	0.36	19.09	61.92	4.27	2.01	2.05	0.89	0.06	8.92	100
CB15BN021	4	R.H.	0.62	0.90	16.34	71.27	0.59	1.91	0.92	0.82	0.05	6.59	100
CB15BN023	4	R.H.	0.35	0.33	15.77	69.75	2.83	1.11	1.42	0.80	0.03	7.61	100
CB15BN017	4	R.H.	0.46	0.84	20.15	63.63	1.81	2.20	1.94	0.92	0.06	8.00	100
Average			0.45	1.02	20.10	61.32	2.06	2.10	4.06	1.00	0.06	7.84
Standard Deviation			0.16	0.38	2.98	6.82	1.65	0.40	6.86	0.12	0.05	1.75

The pattern visible on the scatterplots generated from the major element data appears to corroborate this picture. These plots are shown in Figures 8 and 9. Especially interesting is the scatterplot of MgO/Al₂O₃ vs. K₂O/Al₂O₃, as these ratios could help to distinguish different clay sources. Apart from a Palo Seco outlier from group 3 (CB14BN440) and the TRI-06 clay source, all data points seem to cluster in a more or less elliptical cloud. The samples from group 1 are scattered quite widely apart within this cloud whereas the samples from groups 2, 3 and 4 cluster more tightly together within their respective groups, albeit with significant overlap between the groups. It is important to note however, that the chemical variability displayed in this plot is quite small and the overall differences between the data points are limited. There is significant overlap between many of the clay sources and the ceramic samples, although both ratios tend to be higher in the clay samples than in the ceramics. Some clays (e.g. TRI-11, TRI-12 and TRI-15) do plot outside of the main ceramic cluster.

Figure 7. from left to right and top to bottom: CB14BN486, CB15BN020, CB15BN026, CB15BN021, CB15BN023 and CB15BN017 under XP.

Figure 8. MgO/Al₂O₃ vs. K₂O/Al₂O₃ scatterplot.

3.2.2. Trace elements

Trace elements could aid in comparing the ceramics with potential clay sources. For most trace elements, the clay sources and the bulk of ceramic samples coincide nicely, as illustrated by Figures 10 and 11. There are some exceptions: Ba and Y for example tend to be significantly lower in the clay samples than in the ceramics (Figures 12 and 13). Ceramic samples from the same petrographic group tend to cluster together, but display a significant amount of overlap with the other groups. We can conclude that some of the clays and the ceramic groups are fairly homogeneous in terms of trace element compositions. This could be an argument against clay mixing.

Figure 9. Na₂O vs. MgO scatterplot.

Figure 10. Cu vs. Zn scatterplot. Bulk of samples between 20–50 ppm for Cu and 90–250 ppm for Zn.

Figure 11. Cr vs. Ni scatterplot. Most samples between 50–160 ppm for Cr and 15–80 for Ni.

Figure 12. Zr vs. Ba scatterplot. Most samples between 100–270 ppm for Zr. For Ba, clays between 350–500 ppm and most ceramics 400–1000 ppm.

Figure 13. Sr vs. Y scatterplot. Most samples 90–400 ppm for Sr (some exceptions from groups 1 and 3). Most samples 20–50 ppm for Y with outliers from groups 1 and 4.

4. Discussion

An important point of discussion is the comparison of the information generated from the thin section petrography and the chemical analyses, and the archaeological information about the ceramics.

It was mentioned earlier in this paper that Erin is important in the production of sponge and fine quartz tempered wares. CB14BN459 and CB14BN480 seem to corroborate this and confirm that such wares were also found in Quinam, which is situated approximately 17 km further to the east. Since many of the clay samples were chemically similar, it is impossible to assign a more specific production location on the island. Petrographic and chemical data are both consistent with local (surface) geology, so an origin local to the island is likely. In terms of this surface geology, both sites are situated within a Tertiary and Cretaceous marine stratum, as are all other sites except Red House, St. Joseph and Bontour and all clay sampling locations except TRI-06 and TRI-19/20. Figure 1 in the introduction presents an overview of the locations of the sites and clay sampling locations and the surface geology.

No information about a fabric tempered with metamorphic rocks was found in Boomert (Boomert 2000; Boomert 2013; Boomert 2016). If all investigated Red House sherds are indeed Saladoid, then for the first time in the current study there is a clear difference between the typological assignment of the sherds and their petrography, as these samples were petrographically very distinct from the other Saladoid material under consideration here. It could suggest that the Saladoid pottery production was very localised, since the surface geology map shows that Red House and St. Joseph are both situated close to a stratum with low to intermediate metamorphic rocks, suggesting local sand was used as temper and/or local clays were used. The only clay source from this stratum which was analysed, was TRI-06. However, this sample was taken quite far from the sites, which may be an explanation why the trace element compositions do not closely match. In fact, TRI-06 is an outlier for most major and trace elements. There is significant overlap with other clays sampled from the island suggesting a local origin, while the metamorphic inclusions strongly hint at production in the north of the island.

Fabric group 3 (grog) was found on nearly all sites studied and can probably be assigned to the Saladoid type as Boomert (Boomert 2000) points out that grog temper occurs on the island during the Saladoid period. This may also explain why this group was represented in thin sections from all sites studied here except Bontour and St. Joseph, as these ceramics displayed a wide distribution (Boomert 2000, 127–169). The trace element composition is consistent with an origin local to the island.

Due to the fact that the compositional ranges for both major and trace elements are fairly narrow for all ceramics and clays investigated in this study, it may prove possible in future research to use this typical “Trinidad” signature when executing inter-island comparisons.

The fact that all Trinidad ceramics studied here are consistent with a local signature is interesting. Nearly all of the other pre-colonial ceramic assemblages from islands of the Lesser Antilles which have undergone provenance determinations, showed at least a mixture of locally produced and imported wares (Crock 2008; Fitzpatrick, Carstensen, and Marsaglia 2008; Isendoorn, Hofman, and Booden 2008; Pavia, Marsaglia, and Fitzpatrick 2013; Lawrence, Marsaglia, and Fitzpatrick 2016; Scott 2018). The Saladoid may have found its way into the other Lesser Antillean islands from its origins in the American mainland trough Trinidad and stylistically shows similarities across all the islands of the Lesser Antilles where it is attested (Hofman and Reid 2014; Boomert 2014), but it is currently still unclear how deeply this uniformity permeates into the very fabrics. This would form interesting source material for further study, as would a comparison with other artefact types. Boomert (2013, 144) notes in this respect that at least ornaments of semi-precious rock fragments found on both Trinidad, Tobago and the Windward islands point towards intensive interaction during Saladoid times. For later pre-contact periods, the apparent lack of interaction with the rest of the Lesser Antilles is less surprising since post-Saladoid Trinidad typologically shows larger affiliations with the Venezuelan mainland (Boomert 2014). This is also evidenced by the fact that the Barrancoid, Arauquinoid and Mayoid series originally had mainland origins and that e.g. during the Arauquinoid period, exchange objects from the mainland can be found on Trinidad and that the main tempering material during the Mayoid, caraipé (burnt tree bark), is typical of the Lokono in the Guianas (Boomert 2014, 337–338).

5. Conclusion

Through a combination of information from literature, thin section analysis and chemical analysis, it was possible to sketch a clearer picture on the manufacturing of Trinidad’s pre-colonial ceramics. In this paper, four main ceramic groups were identified in thin section, of which all groups corresponded to groups described previously in the literature. Moreover, it was seen that apart from the metamorphic group, which appears to be local to the north of the island to sites such as Red House and St. Joseph, samples from the other groups appear on multiple sites. It is, however, problematic that no more detailed chronology is available as this makes it very difficult to decide whether for instance the presence of both group 3 and group 4 ceramics in Red House is due to certain contemporaneous consumption patterns, or simply signifies a change through time. Despite restricted sample size, the combination of petrographic and chemical analyses together with information from surface geology suggests that a production local to the island is probable for all ceramics studied. Moreover, the “Trinidad” chemical signature defined here will be useful for further archaeometric investigations of Caribbean ceramics and clays since it fills a blank spot in archaeometric ceramic data from the Lesser Antillean region.

Acknowledgements

The results presented here are an outcome of the CARIB project “Caribbean connections: cultural encounters in a New World setting” financed by the Humanities in the European Research Area (HERA grant number: 1133), of the NEXUS1492 ERC-Synergy Project from the European Research Council under the EU’s 7th Framework Programme (FP7/2007-2013) / ERC grant agreement n° 319209 and the NWO grant “Island Networks: modelling inter-community social relationships in the Lesser Antilles across the historical divide (AD 1000-1800)” (NOW grant # 360-62-060). Furthermore, we would like to extend our thanks to Daan Isendoorn for the collection of some of the ceramic and clay samples and to Herman Nijs, Steven Luypaers and Elvira Vassilieva for preparation of thin sections and ICP-OES analyses. Our gratitude also goes out to Arie Boomert for his valuable comments on the archaeology of Trinidad.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Anneleen Stienaers is PhD researcher at KU Leuven and Leiden University studying continuity and change in the ceramic repertoire of the Caribbean from an archaeometric viewpoint. Prior to her current position, she completed a BA in Archaeology at KU Leuven (2015) and an MSC Experimental Archaeology at the University of Sheffield (2016) with a dissertation titled: “Concerning crucibles: the influence of ceramic containers on Egyptian and Roman soda glasses.”

Bert Neyt was a Post-doctoral researcher at KU Leuven until 2016, where he compiled a database of petrographic and chemical analyses of pre-colonial ceramics of the Lesser Antilles.

Corinne Hofman is professor of Caribbean archaeology at the Leiden Faculty of Archaeology and the Royal Netherlands Institute of Southeast Asian and Caribbean Studies (KITLV) of the Royal Dutch Academy of Sciences (KNAW). After obtaining a BA degree in art history and archaeology at the Vrije Universiteit Brussels (VUB), she completed her MA in pre-Columbian archaeology in 1987 and her PhD in 1993 at Leiden University with a research focus on the Caribbean. Since then her research and teaching focusses on the archaeology and indigenous history of the Caribbean.

Patrick Degryse is professor at the department of Earth and Environmental Sciences and director of the Centre for Archaeological Sciences at the Katholieke Universiteit Leuven (Belgium), and professor of Archaeometry at the faculty of Archaeology in Leiden University (the Netherlands). His main research efforts focus on the history and use of mineral resources in ancient ceramic, glass, metal and building materials production, developing geochemical techniques for characterisation and provenancing.

Additional information

Funding

This work was supported by European Research Council: [Grant Number 319209]; Humanities in the European Research Area: [Grant Number 1133]; Nederlandse Organisatie voor Wetenschappelijk Onderzoek: [Grant Number 360-62-060].

Notes

1 All calibrated dates with 95.5% confidence interval.