During the summer of 1999, chert samples were collected in the field from locations scattered across most of the outcrop extent of the Prairie du Chien formations in southeastern Minnesota and southwestern Wisconsin. Figure 3 shows the twenty locations at which geologic samples were collected. The majority of sample locations are located within the "driftless" area near the Mississippi River, because this is where the best outcrops are available. A few locations were also sampled in the far western portion of the Prairie du Chien, near the Minnesota River Valley.

In order to obtain a set of samples that will result in a data set most accurately representing an overall rock formation, one must have some understanding of the extent and kind of variation within the formation. Ironically, this can only be confidently achieved by first collecting samples and analyzing them. Without this recursive step, however, a number of assumptions or expectations can be made to guide the sampling strategy. Most simply, the number of samples required for a study will increase with the amount of variation present within the sampled body and also with the degree of desired statistical certainty.

Luedtke and Meyers (1984) have addressed this issue, looked at some specific examples, and come up with some generalized models. Extending earlier work done by Luedtke (1976), their statistical examination of samples from nine Midwestern chert types presents the unsurprising conclusion that in general, the amount and extent of variation within a chert-bearing formation increases along with the size of the area sampled. They also produced models describing six potentially different ways in which the trace elements within a formation might vary. The possibilities range from plain homogeneity to simple uni-directional variance to complex, unpredictable variations (see Figure 4). This last one, their model #6, is the most likely to be an accurate expectation for a geologic formation; this has both advantages and disadvantages. Fortunately, it means that different portions of the formation should have distinct signatures; unfortunately it also means that a dense sampling strategy is probably necessary to produce an accurate understanding of the pattern of the variations.

If trace elements behave according to one of the more simple distributions, then a few well-chosen sample locations across the formation would be sufficient to understand the pattern. However, the trace elements within the Prairie du Chien Group might just as likely behave in a manner similar to model #6 in Figure 4. Gaining an accurate picture of something this complex requires a more dense sampling strategy; in fact it should be as dense as feasible since the spatial scale of elemental variations is unknown.

The search for suitable sampling sources was influenced in two ways: First, it was desired that samples might be obtained from geologic sources near archaeological sites that archaeologists have interpreted to have been the sites of chert procurement by prehistoric Native American groups. (Such locations are frequently referred to as "quarry sites", but this is often an inaccurate term since actual quarrying of material from its bedrock source probably seldom or never took place when there were plenty of useable chert nodules naturally eroded out and littering the surface or stream beds.) There is an abundance of such raw material procurement sites identified in the bluff and coulee region of Minnesota, Wisconsin, and Iowa. In an archaeological reconnaissance survey conducted in Mill Coulee and adjacent uplands near the town of Prairie du Chien, WI, Theler (1981) located 95 previously unknown sites; 81 of these he interpreted as "lithic processing stations" where the procurement of Prairie du Chien chert was the principle activity. (Such lithic procurement and processing sites are identified by the high concentration of debitage from the earliest stages of the lithic reduction process.) There was nothing particularly out of the ordinary about the area which Theler surveyed, and it gives an idea of how commonplace chert procurement sites may be all along the bluffs and coulees near the Mississippi River.

The archaeological sites near which geologic samples were collected include the St. Croix Access site near Stillwater, MN, with Archaic and Woodland occupational components (Hoffman and Myster, 1992); the Dutchman Coulee and King Coulee sites near Wabasha, MN; the Gran Grae site near Prairie du Chien, WI, with Late Archaic and Woodland components (Arzigian, 1981); and the Chiparoo site west of Eau Claire, WI, with Late Archaic and Late Woodland occupational components (Hawley et al., 1998). In order to obtain a geochemical fingerprint for the majority of the Prairie du Chien Group, however, a number of sample locations were chosen that were not associated with any known archaeological sites.

It should be noted that it was not always possible to obtain chert samples from outcrop sources; it was sometimes necessary and/or desirable to acquire samples from stream deposits rather than removing them from bedrock. This is true for the sample locations near Mankato, and those near the Stillwater and Wabasha archaeological sites (see Figure 3). This reflects an important aspect of the archaeological side of this study. The prehistoric people whose cultures we are trying to understand probably only rarely obtained chert from bedrock sources; rather they acquired chert from lag or stream deposits where it could be gathered with much less effort.

Acquiring chert from stream deposits, however, raises the question as to whether or not samples from stream deposits (which are more weathered and perhaps have been transported some distance) will differ geochemically from outcrop samples. To address this concern, there was one sample location at which chert samples were collected from both outcrop and stream deposits. This was at sample location #6 near Winona, Minnesota. The location consists of a small, deeply-incised intermittent stream which alternated between a cobble/gravel and a bedrock-floored stream bed, with bedrock outcrops flanking the stream. One set of samples was collected from the stream-deposited cobbles, and another set was collected from the bedrock outcrops. The geochemical data from these two sets of samples therefore can be compared to address the question of weathered versus unweathered samples.

A total of ten chert samples was collected from each sample location. These ten samples were gathered to represent the maximum vertical, horizontal, and textural variability present at each location. A larger number of samples would have been better for this purpose, but that factor had to be weighed against the analytical constraints. The samples were all labeled, and a brief description and a geologic-context notation were made for each one. The samples were then taken to the University of Minnesota, Duluth, Archaeometry Laboratory for sample preparation.

     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

The views and opinions expressed in this page are strictly those of the page author.
The contents of this page have not been reviewed or approved by the University of Minnesota.