Read an Excerpt

Ceramic Petrography and Hopewell Interaction

The University of Alabama Press

Methodology

Thin Sections and the Petrographic Microscope

Petrography is a venerable geological technique that provides reliable identifications of minerals and rocks. In this role it is well suited for the analysis of pottery because, as a mixture of mineral and rock fragments, most of which are natural inclusions but some of which may also be intentional human additives (i.e., temper), pottery can effectively be treated as if it were a metamorphosed sedimentary rock (Williams 1983:302). Knowledge of these minerals and rocks is valuable for archaeological interpretations concerning the production, use, and exchange of pottery vessels, all topics that will be addressed in the following chapters.

One liability of the technique is that it is destructive, requiring the sacrifice of a slice of a sherd that is then prepared and mounted on a standard glass petrographic slide. The glass slides measure 27 mm x 46 mm, which thus constrains the size that a thin section can be. Normally, sherds are sliced at right angles to the vessel wall, which means most thin sections measure less than 10–15 mm in breadth (depending upon the thickness of the vessel walls) and can be no longer than the maximum length of the slide. Sherds can be cut transversely, i.e., parallel to the vessel walls, thus producing a larger thin section but at the sacrifice of a greater portion of a sherd. For small sherds—say smaller than a quarter-dollar in maximum diameter—little if any of a sherd will survive. But, for larger sherds, a significant portion may be returned to the original donor with only minor damage or retained as a "voucher" for future reference.

To prepare a sherd for mounting on a glass slide, it is first cut to create a flat surface that is further refined by polishing. An epoxy is then used to attach the polished sherd to a glass slide. After the epoxy cures, the sherd is again cut parallel to the initial cut, resulting in a "thick section" (perhaps 1 or 2 mm thick) now permanently attached to the slide. In the final step, one requiring considerable experience and skill, the thick section is ground down to a standard thickness of 30 microns, i.e., .03 mm. The reason for this latter, important step is that the behavior of light passing through a mineral (which is essential for making identifications) is dependent upon two variables, the thickness and the character of the mineral. Thus, in order to facilitate mineral identification, the thickness of the thin section is standardized. All literature describing the optical properties of minerals (e.g., Phillips and Griffen 1981) are premised on 30 microns as the thickness of thin sections. Once mounted, the thin section is usually covered with a protective cover glass and is now a permanent record.

All of this care in the preparation of a thin section is required for its effective analysis on the horizontal, rotatable stage of the petrographic microscope. Unlike biological microscopes, which illuminate specimens with incident light from above, the light source for petrographic microscopes is below the stage from where it is transmitted up through the thin section to the eye of the observer.

In addition, there are several accessories to a petrographic microscope that are essential for its effective use. Especially important are two polarizing lenses (also called nicols after their inventor, William Nicol), one permanently mounted below the stage (the polarizer) and another located above the stage (the analyzer) that can be inserted or withdrawn by the analyst.

In observing the thin sections the analyst works both with the analyzer withdrawn from the field of view (i.e., the light passing through the thin section derives solely from the polarizer and is referred to as plane-polarized light (PPL), because it is vibrating in a single plane dictated by the polarizer) and with the analyzer inserted, i.e., under crossed polars (commonly referred to as crossed nicols and abbreviated XP). When viewing thin sections under plane-polarized light, the natural colors of minerals are visible, which is extremely informative. By contrast, under crossed polars significant optical properties appear (that are complicated to explain) that involve major color changes (so-called interference colors) that are also important for identifying most minerals. As will be seen in later chapters, photomicrographs of thin sections will be shown under both plane and crossed polarized light, for each has special, complementary properties (Figure 1.1).

A second liability of petrography is that it is incapable of identifying individual clay minerals because of their small grain sizes (i.e., <.002 mm). On pottery thin sections, clays appear as a homogenous matrix whose composition can be identified no further. Thus, the focus of petrographic analysis is upon individually visible grains of silt size (>.002 mm) and larger.

A third characteristic (liability?) of petrography is that it deals directly with the physical properties of rocks and minerals, thus does not provide chemical compositional data. Inferences are sometimes possible as to the chemical makeup of a grain, but this is based solely upon the physical properties observed, not by direct determination, as, for example, is possible with neutron activation or scanning electron microscopy.

The petrographic analysis of pottery came to worldwide attention through the pioneering work of Anna O. Shepard (1936, 1939, 1956). In two studies involving pottery from prehistoric villages in New Mexico—Pecos Pueblo east of Santa Fe (Shepard 1936) and Pueblo Bonito in Chaco Canyon (Shepard 1939, 1954)—Shepard recognized much of the local pottery contained nonlocal igneous rock tempers. She suggested that such pottery was produced at specialized workshops elsewhere, and imported, views that were counter to the prevailing archaeological wisdom of the time that saw such Pueblo communities as economically self-sufficient. These early studies focused entirely on the qualitative identification of tempers, seeing them as exotic based upon the known geological distribution of the rocks involved. One of the arguments used against her views was the possibility that the rocks, not the vessels, were imported—the pots-versus-temper-importation issue that Neil Judd, the excavator of Pueblo Bonito, felt "still hangs in midair" (Judd 1954:235) after an exchange of letters with Shepard.

Always the cautious scholar, Shepard regarded her observation of temper similarities between Chaco Canyon and the Chuska Mountains around 70–80 km to the west as "only circumstantial evidence, not proof" and suggested that a more comprehensive analysis be conducted comparing the pottery from the two areas that involved various stylistic considerations and consideration of the type of clays (Shepard 1954:236–237). I mention this situation because it was this expressed need by Shepard for more comprehensive analysis than temper-only observations in petrographic analyses that inspired my effort to introduce a quantitative component to such studies. Later, I was privileged to be able to apply the quantitative approach I was developing to Shepard's very own thin sections from Pueblo Bonito and the Chuska Mountains and, I believe, confirm her importation-from-the-Chuskas hypothesis (Stoltman 1999a).

Point Counting: A Quantitative Approach to Thin Section Analysis

For this study, all thin sections were analyzed using a basic point-counting procedure as described in several earlier publications (Stoltman 1989, 1991, 2001). The goal is to characterize the physical composition of thin sections in both quantitative as well as qualitative terms for use as independent evidence, along with stylistic and contextual considerations, to provide a robust data set for addressing issues of ceramic production and exchange.

The procedure of point counting is analogous to superimposing a rectangular grid over a thin section and recording the minerals that appear at every point of intersection on the grid. The actual procedure, however, is different and less prone to error. It requires a petrographic microscope specially equipped with two properties: (1) a measuring eyepiece with a crosshair and (2) a stage attachment that permits the movement of the thin section back and forth at fixed intervals across the stage beneath the crosshair. With a microscope equipped in this way, the analyst does not have to keep track of where in the grid observations are being made, but only concentrate on identifying whatever appears beneath the crosshair at each stop of the thin section as it is moved across the stage. The main goals of point counting are to provide unbiased estimates of the constituents of a thin section expressed either as a percentage of volume or as a percentage of individual grains.

At the outset it must be emphasized that point counting (referred to as modal analysis in the geological literature, e.g. Chayes 1956; Bayly 1960), is a sampling procedure that is subject to a number of sources of error. Perhaps most important is the issue of sampling error, i.e., the relationship between the sample and the population from which the sample was derived. In dealing with pottery, a thin section is clearly a sample of a vessel, which in turn is a sample of a ceramic assemblage, the latter normally being the target population about which compositional information is being sought. One comforting thought about how reliably a pottery thin section may represent the composition of a vessel is the normal practice of potters to carefully and extensively knead together the ingredients that go into the formation of a vessel. In an empirical test of the proposition of vessel homogeneity, four thin sections from a single vessel were point counted two times each (Stoltman 1989:150–153). The greatest observed compositional difference from the grand mean of the eight counts (presumably the best estimate of the vessel composition) was 2.3 percent (Stoltman 1989:153), which is consistent with the view that thin sections of pottery vessels can usually be accepted as unbiased indicators of vessel composition. Nonetheless, the possibility that a thin section derives from a portion of a vessel that is atypical in some way (for example, has a unique grain or was poorly mixed) must always be kept in mind.

Counting error is another possible source of error in point counting. This source of error is defined as the differences between two or more counts of the same sample (Chayes 1956:59–60; Stoltman 1989:152). Counting error may also be viewed as a measure of precision, which "refers to the size of deviations from the mean m obtained by repeated application of the sampling procedure" (Cochran 1963:16). In an effort to evaluate this error source in the point counting procedure used in this study, 25 pottery thin sections were point counted 2 times each, with 6 of these counted a third time (Stoltman 1989:153–155). The maximum amount of counting error for any pair or triad of observations was less than 5 percent so that the level of precision obtainable with this procedure can be expected to be 95 percent or better (Stoltman 1989:153–155).

Point counting pottery sherds is not precisely analogous to point counting geological samples. It is common for geologists to count several hundred to over one thousand points, but this involves using multiple thin sections derived from rocks that may cover several square kilometers. By contrast, archaeologists are hard pressed to obtain more than a single sherd from a vessel, which means that the thin section area available for counting is severely limited. Accordingly, certain accommodations have to be made.

A thin section may reasonably be regarded as a random sample of a pottery vessel (really, there is no alternative), but the actual points counted cannot be truly random and at the same time occur in large enough numbers over the delimited area of a pottery thin section to have any reasonable expectation of attaining a statistically reliable sample. Thus, rather than attempting a random sample, the point counting procedure used here following Chayes (1956:11) is best described as systematic sampling, which he feels "will usually be more precise than a simple random one" (Chayes 1956:12).

Through trial and error, a compromise-sampling scheme was devised that yields consistent results. A sampling interval of 1 mm is used. It has proven to be the largest practical interval for most pottery thin sections, yet is small enough to ensure a minimum count of 200 points in most cases. In the early years of this study the normal practice was to count until the total area of a thin section had been covered or until 200 points, excluding voids (i.e., cracks or flaws), had been reached, whichever came first. In more recent years a minimum of 200 points was always counted, reversing the slide if necessary and counting in the opposite direction in order to reach the 200 points.

Experience has shown that by employing the 1-mm counting interval, it is possible to compile simultaneously both volumetric and individual-grain data for pottery thin sections. For such variables as relative amounts of sand or silt or temper, etc., volume of such inclusions is the most appropriate measure. It is here that the so-called Delesse relation applies, i.e., that the area of a mineral provides an unbiased estimate of its volume in a thin section (Chayes 1956:4–15). The 1-mm interval allows the analyst to keep track of any multi-count grains that occur. Indeed, this must be done for volumetric estimates to be valid. With a smaller counting interval, pottery thin sections with multiple grains of coarse sand and even gravel sizes (a common property) will become virtually impossible to count reliably. By contrast, individual grain counts may also be recorded simply by keeping track of multi-count grains and counting them as one when, for example, compiling size indices for sand and temper.

The result of this approach to point counting is a physical characterization of each thin section in terms of the percentages of silt-size inclusions plus the species, sizes, and percentages of mineral inclusions of sand size and larger. Since individual clay particles are not identifiable in thin section, their presence is recorded simply as "matrix." The point counting was routinely conducted "blind," that is, without prior knowledge of the samples other than their thin section number. The data for individual thin sections were neither tabulated nor compared until the analysis of each data set had been completed.

Six indices are used to express the results of the point counting: (1) bulk composition; (2) a mineralogical index that accompanies bulk composition; (3) body; (4) paste (these four are volumetric measures); (5) sand-size index; and (6) temper-size index (the latter two are grain-count measures). Bulk composition and its accompanying mineralogical index plus the sand-size indices are used to characterize those vessels that lack visible or reliably identifiable temper, i.e., are untempered or sand tempered (for the latter it is normally impossible to determine objectively which sand grains are natural inclusions and which are human additives).

Bulk composition is a volumetric measure comprised of the relative percentages of three variables, matrix/clay, silt, and sand (always understood to include gravel when present). These are all purely size grades as follows: matrix/clay=<.002 mm; silt= visible grains ranging in size from .002 mm to .0624 mm; and sand/gravel= grains larger than .0625 mm in maximum diameter. As a complement to this index, a sand-size index is also recorded for each vessel. This is devised by assigning a value within the following ordinal scale (based on the Wentworth Scale; e.g., Rice 1987:38) to each sand-sized (or larger) grain encountered in point counting based upon its maximum diameter:

1. Fine=.0625–.249 mm

2. Medium=.25–.499 mm

3. Coarse=.50–.99 mm

4. Very coarse=1.00–1.99 mm

5. Gravel=2.00+ mm

The size values for each of the grains were then summed and divided by the total number of grains counted, thus providing a mean "sand-size index," which ranges between one and five for each thin section.

In addition a mineralogical index will be utilized for vessels that are characterized by bulk composition. This index is comprised of three variables: percent monocrystalline quartz; percent polycrystalline grains; and percent FMM (i.e., feldspars, mafics [i.e., dark minerals such as amphiboles and pyroxenes], and micas). This index is derived from counts on all sand and larger grains whose mineralogy was recorded during point counting. Care is required in using this index because the number of grains counted for such vessels typically range between 50 and 100. Thus, for such small samples the relative percentages among the three classes are prone to sampling error. If used cautiously, this index is useful evidence to enhance our understanding of the mineralogical makeup of thin sections characterized by bulk composition.

---

Excerpted from Ceramic Petrography and Hopewell Interaction by James B. Stoltman. Copyright © 2015 University of Alabama Press. Excerpted by permission of The University of Alabama Press.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---