INTRODUCTION AND BACKGROUND
Provenance, according to archaeological definition, is the study of artifacts with the intended purpose of determining their geographic origin. By comparing the characteristics of materials with known origins with samples of an unknown origin, the unknown samples may be attributed to particular sources. This process usually involves examining the physical characteristics of the material from which the artifact was made. This information can be of immense use to archaeologists, anthropologists, and historians in understanding past cultures. Where an artifact is discovered may be kilometers or hundreds of kilometers from the location where it was originally manufactured. In between there may have been a long history of use or trade. Successful interpretation of provenance is necessary for the recreation and understanding of migration patterns, trade routes, interaction networks, territory size, economic systems, and procurement procedures. It is this premise which has encouraged the development of provenance studies in archaeology.
Archaeological provenance studies have developed with close links to geology and geochemistry. This is because the majority of archaeological materials preserved in the archaeological record have their origins as geologic materials; stone and pottery are by far the most common, followed by metals and marine shells. In some ways this is very fortunate, since many geologic materials have distinct geographic distributions. An investigator may be able to identify the exact rock formation from which a particular stone artifact had its origin or the source area of the sediments used in making a ceramic item.
Methods of Provenance
In the case of stone artifacts, archaeologists have long attempted to discover geologic provenance using the most simple of methods looking at the color, texture, and other macroscopic visual properties of the stone. Such an analysis has clear advantages in time, cost, and simplicity. This is a distinctly limited method, however, due to the enormous amount of overlap in visual characteristics between various geologic materials, including cherts. The most basic visual property, color, can also be the most misleading, since it is affected by grain size, texture, and inclusions. Thus macroscopic identification is often misleading and seriously inadequate when dealing with chert.
This problem is made more complicated because most archaeologists do not have geologic training or extensive geologic knowledge of the regions in which they conduct excavations. Having seen only the raw materials recovered from archaeological excavations and not having visited many, if any, outcrops, there is likely more variability within the full range of a geologic source than the archaeologist is aware of.
In a provenance study on archaeological cherts from southern Illinois, Spielbauer (1984) found serious errors in previous, visually-based sourcing classifications of what were thought to be three easily-recognizable, well-defined chert varieties. He found that the traditional macroscopic typological techniques lacked the consideration of the full range of macroscopic variations, the utilization of standards for color comparison (such as the Munsell Color Schema), and the often-encountered problem of artifacts too small to show the full range of physical variation and characteristics for identification.
The problems inherent in macroscopic visual analyses do not make it a useless technique, however. It may be used for non-quantitative studies that do not require high accuracy, or it may be combined with more quantitative methods for more in-depth studies. Malyk-Selivanova et al. (1993) conducted a complex provenance study of chert artifacts and sources in northwestern Alaska using several geochemical techniques, petrographic, and macroscopic analyses. They confidently correlated artifacts to sources using both chemical and petrological information Despite finding visual characteristics alone to be unreliable, the authors stated that putting together all the analytical information was necessary for a successful determination of provenance.
The next step beyond simple macroscopic analysis might be to examine the artifacts microscopically, perhaps in thin-section. By examining the chert samples with a petrographic microscope, significant microfossils or mineral inclusions may aid in provenance. While providing more information, this also has distinct limitations. Not least of these limitations is the destructive nature of making a thin-section. Archaeologists will not allow any but the least significant of artifacts to be sectioned for microscopic analysis. Also, the monotonous and fine-grained nature of most cherts leaves few potentially distinguishing characteristics to see under the microscope.
Perhaps the most useful methods for geoarchaeological provenance are those involving geochemical analysis. The analysis of trace-element content provides an accurate and replicable means of sourcing because geologic materials often have diagnostic chemical properties. Also, geochemical analyses provide a far more objective means of sourcing geologic materials than visual techniques. Results obtained by geochemical analysis can be published so that other researchers can judge the same or similar data to see if they arrive at similar conclusions. With visual techniques, however, it is impossible for the reader of provenance literature to assess the validity of the authors conclusions.
Not surprisingly, over the past few decades the study of geochemical provenancing has grown increasingly popular and more accurate with the development of physical methods of analysis. Such methods include Mass Spectrometry, X-Ray Fluorescence, and Neutron Activation Analysis. The great potential of these research techniques to aid in archaeological studies was understood quickly and applied by some researchers. In particular, some of the earliest intensive studies of this sort were conducted on prehistoric flint mines in the U.K. using Atomic Absorption Spectroscopy (Sieveking et al., 1972) and Neutron Activation Analysis (Aspinall and Feather, 1972).
Neutron Activation Analysis
Instrumental Neutron Activation Analysis (INAA) is extremely well-suited to trace-element studies, since it is highly sensitive to almost all elements of interest and returns data with relatively high precision. It is a technique which permits simultaneous analysis of a large number of elements in a sample, particularly the heavy trace elements which are often crucial for the chemical fingerprinting of geologic materials. Also, objects can be analyzed by INAA using minimally-destructive methods, using either a whole object or a very small amount of material removed from it. This is desirable when investigating irreplaceable archaeological artifacts. In addition to offering data on a broad range of elements, INAA also provides sensitivity down to the detection of elemental concentrations of parts per million or even parts per billion. For these reasons, INAA has become a popular method of analysis for geologists and archaeologists for items ranging broadly from marine shells (e.g. Claasen and Sigmann, 1993) to pottery (e.g. Bieber et al., 1976), cassiterite (Rapp et al., 1999) and flint (e.g. de Bruin et al., 1972).
In the Instrumental Neutron Activation Analysis process, samples are subjected to a beam of neutron radiation for a precise period of time. The incident neutrons react with nuclei in the sample, producing compound nuclei with increased atomic mass but no immediate change in atomic number. Each heavier nucleus is now in an excited (unstable) state, and is referred to as a radionuclide. After a period of time determined by the unique half-life for each radionuclide (from a fraction of a second to many years), the unstable nucleus decays via the emission of a Beta particle and a gamma ray. Due to the Beta emission, one neutron in the nucleus is transformed into a proton, increasing the atomic number of the nuclide but maintaining the same atomic mass. The wavelength (and therefore energy) of the emitted gamma ray is distinctive to the isotope that produced it.
An example of this process can be made with sodium: 23Na interacts with an incident neutron, producing 24Na (an unstable radionuclide). After a period of time, the 24Na emits a Beta particle and a gamma ray of specific wavelength. After undergoing this decay, the nuclide now becomes 24Mg. The emitted gamma-ray interacts with a detector, revealing the presence of sodium in the sample.
The gamma-ray detectors used for INAA are typically germanium semiconductor detectors linked to a computer-controlled multi-channel analyzer. Resolution and sensitivity are the most important factors for accurate element detection. The resolution of the detecting instrument will determine its ability to distinguish among isotopes with closely spaced gamma-ray wavelength peaks. The sensitivity of the instrument affects how small a concentration of isotopes can be detected and how accurately it will determine the concentration of isotopes in the sample. No matter what the sensitivity, the analysis can only be considered qualitative unless the detected gamma-ray peaks are calibrated with a sample of precisely known composition.
To accurately calculate the concentration of elements in an unknown sample, the unknown is irradiated along with a known standard (preferably of similar composition), and the two are both measured using the same detector. Elemental concentrations are then calculated using a formula incorporating information on the radiation activity of the sample, atomic density of the sample, neutron flux of the reactor, the reaction cross-section, the efficiency of the detector, decay constants for the isotopes, and elapsed decay time. The ratio of calculated values between the standard and the unknown sample will then provide a quantitative number for the concentration of elements in the sample. Accuracy by these techniques usually range between 1 to 10 percent of the reported value. The use of common standards by different analytical laboratories permits the direct comparison of data obtained by different researchers at different institutions and times.
The INAA method has some limitations, including an inability to detect lighter elements such as hydrogen, carbon, nitrogen, oxygen, and silicon. These elements are not observed, mainly because they do not create detectable radioisotopes. This is not a serious problem, however, because in general these lighter elements are either not present in abundance or are too ubiquitous in geologic samples to be of use in provenance studies.
As can be seen from the methodology outlined above, the accuracy and quality of data generated by INAA is dependent upon a number of factors. A broad understanding of the process and limits of INAA is necessary to fully understand the validity, accuracy, and potential problems of data generated in this way. Overall, most researchers using this technique have found INAA to produce good-quality, accurate data, and it is well-suited to geochemical provenance studies.
Provenance of Chert
In the Midwest attention has focused in particular upon the geochemical provenance of cherts. This is because chert was one of the most abundant, utilized, and enduring raw materials in prehistory. There are a large number of geologic formations that yield varieties of useable chert, and (unfortunately for archaeologists) most of these different cherts have a great deal of overlap in visual characteristics. This often makes it difficult if not impossible to distinguish them reliably without geochemical analysis. For these reasons, a number of studies in the past few decades have focused attention upon the use of INAA to determine the provenance of Midwestern and Great Plains cherts (e.g. Hoard et al., 1993; Luedtke, 1979; Spielbauer, 1984).
The use of geochemical sourcing techniques has revealed some erroneous assumptions that had become entrenched in the archaeological literature concerning the prehistoric sources of raw materials. In many of these cases, attribution of lithic materials to particular sources had been over-simplified, due in part to a reliance on visual classification. For instance, Morrow et al. (1992) focused an analysis upon cherts of southwestern Illinois and found that the broadly accepted characterization of chert sources did not reflect the true complexity of the matter. Similarly, Boszhardt (1998) identified a number of previously unrecognized sources of lithic material in western Wisconsin and southeastern Minnesota that almost certainly contributed stone for prehistoric toolmakers.
Throughout almost all previous provenance investigations, however, the attempt has been nearly exclusively focused upon differentiating between cherts from entirely different geologic formations and not upon the possibility of distinguishing between geographically separate sources within individual formations. Luedtke (1978) points out that such intra-formational variations likely have not been noted because most provenance studies have analyzed only small numbers of representative samples (up to a few dozen or so) from each formation. A start was made by Luedtke (1978) and Luedtke and Meyers (1984) into investigating intra-formational provenance. Their studies showed some promising results, suggesting that there are systematic geochemical variations within chert-bearing formations; therefore intra-formational provenance studies may be feasible and useful. Luedtke (1978) also noted that the macroscopic visual variations (such as texture and color) of chert samples often correspond with chemical variations, though not in any simple or straightforward manner.
To continue a more in-depth study into the possibilities of intra-formational provenance, it was decided that the cherts of the Prairie du Chien group might present a good test case. Chert from the Prairie du Chien group was one of the most widely used lithic materials of southern Minnesota and western Wisconsin (Bakken 1997; Morrow 1994), although some archaeologists believe that these cherts were not widely traded outside their outcrop area due to the availability of higher-quality materials for trading (Boszhardt 1998). The large geographic distribution and relatively good outcrop exposure of the Prairie du Chien makes it an excellent candidate, however, for an intra-formational provenance study. Other lithic materials from this area include Cedar Valley Chert, Grand Meadow Chert, and Galena Chert, though none of these have as wide geologic distribution nor as good outcrops as the Prairie du Chien (Bakken, 1997).
Extensive efforts have been invested by a number of researchers into the identification, classification, and provenance of cherts based upon macroscopic visual criteria. For instance, after extensive study of archaeological cherts from the upper Midwest, Morrow (1994) created a classification tree to aid archaeologists in the identification of chert artifacts recovered in Iowa. This classification tree consists of 68 questions, mostly requiring yes/no answers concerning the characteristics of the unknown rock sample. Working through the branches of the question tree into ever finer categories should ultimately lead to a specific name for each chert specimen. Since this classification tree includes Prairie du Chien chert, it was used to test visual classification of the geologic Prairie du Chien samples collected for this provenance study. Brief descriptions of the handsamples can be found in Appendix B.
A sequence of three to nine questions in Morrows classification tree is required to arrive at an identification of Prairie du Chien chert. The questions are based upon physical criteria such as color, pattern (mottling, etc.), texture, luster, translucency, fossils, and mineral inclusions. Morrow notes that color is one of the least-useful attributes due to large natural variation within formations, weathering, patinas, and color changes due to intentional heating (which prehistoric people realized often improves the workable properties of the raw material). Morrow recommends that these tests be applied only to samples 2 cm in size or larger because small samples do not have a full range of diagnostic physical features. This can be quite a problem with archaeological artifacts which are often very small. After the Neutron Activation Analysis, two of the original 206 samples had too little material left for reliable macroscopic description. The classification process was followed for each of the 204 remaining hand samples of Prairie du Chien chert, with the following results:
The most common materials which the Prairie du Chien samples were misidentified as were Warsaw Chert (from the Warsaw Formation in central Iowa) and Maynes Creek Chert (from the Hampton Formation in central Iowa). Based upon the lithic raw material descriptions provided by Morrow, these are cherts which are unlikely to be genuinely confused with Prairie du Chien cherts by a person familiar with these types of chert, but a faithful following of the classification tree often leads to them. Other materials which may be confused with Prairie du Chien chert include Jefferson City Chert (originating from southern Missouri), Scotch Grove Chert (from east-central Iowa), and some varieties of Hixton Silicified Sandstone (from western Wisconsin).
Quite probably, the workers familiar with Midwestern chert types would have more success in the visual identification of Prairie du Chien chert than was seen in this experiment. Nevertheless, it illustrates the difficulties encountered when attempting to conduct provenance studies based upon visual criteria. The visual characteristics and methodology by which a researcher comes to a conclusive identification of raw materials is often vague, complex, and difficult to communicate to others. Despite attempts such as classification trees, it is not a readily quantifiable method, and the validity of conclusions based upon it are not readily assessed by other persons.
Petrogenesis and Geochemistry of Chert
Chert by definition is a rock composed almost entirely of microcrystalline quartz (SiO2). It is generally hydrous and consists of complexly twinned a -quartz grains 1 to 50 m m in diameter (Knauth, 1994). There are several varying, much-debated hypotheses for ways in which the different types of cherts may have formed. Nodular cherts such as those in the Prairie du Chien are the result of the replacement of carbonate (calcite or dolomite) sediments with siliceous material, as evidenced by the presence of remnant ooids and sedimentary structures preserved in the silicified nodules. The geochemical replacement typically occurs in small (<30 cm) nodules which in turn often occur in distinct bedding horizons within the host carbonate rock.
Much research has attempted to unravel the details of the silicification process. The solubility of carbonate increases with decreasing pH, whereas silica has the opposite trend; therefore silicification of carbonate can occur when diagenetic pore waters have a high, decreasing pH. A consensus has never been reached concerning a specific mechanism for the conditions of this replacement process to occur. Common theories invoke a range of factors from the mixing of meteoric groundwaters with silica-rich marine pore fluids (Knauth, 1979) to the oxidation of microbial H2S (Clayton, 1986). In a summary review of this subject, Knauth (1994) points out that whatever the process, the evidence is that silicification generally occurs during diagenesis, before compaction, and (in the case of dolomites) after the inception of dolomitization. Since the silicification process in the nodules is usually complete, cherts are commonly 95% or more SiO2 and contain only small amounts of other elements. Still, the rarer and trace elements which make up the remaining few percent of a chert nodule can be extremely informative.
Given the petrogenetic history, there are several possible sources for the trace elements contained within a chert. Some elements may come from the deposited sediments, which were inherited from the "up-stream" weathered and eroded source rocks. The elements within these sediments may have been removed or added to by dissolved phases within the aqueous environment either during transport or deposition, such as adsorption of components from local seawater onto particle surfaces. Further alterations likely occurred during diagenesis of the sediments, subsequent alteration during exhumation and weathering, and possibly during dolomitization in the case of dolostones. While all traces of carbonate may be removed during the silica replacement process, impurities in the carbonate may be inherited by the chert. Thus the trace element content of a chert is a combination of chemical signatures inherited during the processes of transport, deposition, diagenesis, and alteration. The diagenetic factors are quite likely the dominant control of trace-element content because the elemental abundances during chert formation are largely controlled by the addition of dissolved SiO2 (Murray, 1994), which acts as an extreme dilutant.
Variations in chemistry attributable to the source rocks are most likely to be represented by the Rare Earth Elements (REEs), since these elements tend to be transferred nearly quantitatively into the terrigenous sedimentary component (solid load) and their presence in natural waters is vanishingly small. This is in contrast to Na, Mg, Ca, U, and Rb, which are easily mobilized during weathering and diagenesis. Their concentrations therefore are likely to provide information about these processes (McLennan, 1989). Since the Prairie du Chien Group is composed of carbonate sediments with little terrigenous component, their REE content is likely to be low compared to the latter, more easily mobilized, elements (this turned out to be generally true, as can be seen in the element concentration data in Appendix A).
However, a single geologic formation is unlikely to have a homogeneous trace element content throughout the formation, especially if the formation covers a large geographic area. There are likely to be both vertical and horizontal variations in element concentrations between outcrops. These variations can be the product of a number of circumstances. For instance, the basin in which the sediments were deposited might have included microenvironments of varying chemistry, the sources of sediment being fed to the basin may have changed over time, or postdepositional and diagenetic processes may not have acted homogeneously throughout the geologic body. All of these, or any combination of them, provide reason to believe that it might be possible to distinguish varying geochemical signatures within a formation.
Prairie du Chien Group
The Prairie du Chien Group consists of two sedimentary formations which define the Lower Ordovician stratigraphy for southeastern Minnesota and southwestern Wisconsin; these two formations are the Shakopee Dolomite and the Oneota Dolomite. The group was originally named for the set of excellent exposures near the city of Prairie du Chien, Wisconsin, where the group attains a thickness of over 200 feet (Mossler, 1987). Due to the presence of indicators such as ooliths and conodont faunas, the environment of deposition for these carbonates has been interpreted as a shallow marine to marginal marine setting (Smith et al., 1993). Further interpretation has suggested that these dolomites are carbonate sediments characteristic of the tops of carbonate platforms and banks (Smith et al., 1996).
The lower member of the group is the Oneota Dolomite, which has a gradational contact with the underlying (Cambrian) Jordan Sandstone. The Oneota has been subdivided into a lower Stockton Hill Member (sandy dolomite) and upper Hager City Member (dolomite) by some researchers working in Wisconsin (Davis, 1970; Smith et al., 1996). However, these subdivisions are not readily apparent without very close scrutiny (Mossler, 1987), and identification of this contact in the field is difficult (Davis, 1970). No attempt was made to distinguish the Stockton Hill and Hager City members in this study.
Disconformably overlying the Oneota Dolomite is the Shakopee Formation, which is further divided into two members (see Figure 1). The lower member of the Shakopee is the New Richmond Sandstone, ranging in composition from sandy dolostone to quartz arenite (Mossler, 1987; Smith et al., 1996). The upper member of the Shakopee is the Willow River Dolomite. Since the New Richmond Sandstone contains little or no chert, further references made to the Shakopee Formation indicate the Willow River Dolomite member. Whereas the Oneota is typically composed of thick to massive beds, the Shakopee is commonly very thin to thinly bedded. This distinction sometimes provides a useful means of distinguishing between the two in outcrop (Mossler, 1987). The Shakopee Formation is unconformably overlain by the (Upper Ordovician) St. Peter Sandstone, though this upper contact is rarely exposed.
Figure 1: Simplified stratigraphic column, after Mossler 1987.
Both the Oneota and Shakopee formations contain abundant chert nodules, are characteristically oolitic, and are sparingly fossiliferous. In fact, the chert from these two formations is visually indistinguishable and has always simply been grouped as "Prairie du Chien" by archaeologists. An assumption is that chert from the two formations also has a great deal of geochemical overlap, and can be considered to be from one formation. This assumption is addressed and tested later by the analysis of the geochemical data from both formations.
The Prairie du Chien Group crops out over a broad portion of southeastern Minnesota and southwestern Wisconsin (see Figure 2). The greatest concentration of outcrops, however, is within the "driftless area" near the Mississippi River. This influenced the distribution of sample locations, as will be discussed later. Samples were gathered from as large a portion of the Prairie du Chien as was feasible for the scope of this study.
Figure 2: Approximate outcrop extent of the Prairie du Chien Group in Minnesota and Wisconsin. Cross-hatching shows areas of good exposure; single-hatching shows areas of sparse outcrops. Dashed box indicates area shown in Figure 3.
NEXT PREVIOUS TABLE OF CONTENTS
Intro and Background Fieldwork Sample Prep Data Analysis PCA Correspondence Analysis Stepwise DA
Discriminant Analysis More PCA Element Trends Conclusions Bibliography Appendix A: Part 1 Part 2 Part 3 Part 4 Part 5 Appendix B Appendix C