«By James R. Keron Graduate Program in Anthropology Submitted in partial fulfillment of the requirements for the degree of Master of Arts Faculty of ...»
Drumholm Site Observations The Drumholm site is a large Middleport Substage village located west of London on Oxbow Creek. It has been investigated several times by Robert Pearce (1996) and is known entirely through surface collection. Pearce recognized eleven middens on the site in his thesis and conducted a CSP on the site in 1994. This CSP includes the entire site area. At earlier times, material was recovered and was cataloged by midden number.
This material was added to the CSP for the purposes of this analysis. This procedure has one unfortunate drawback in that the samples from the non-midden areas are too small for statistical analysis. Consequently, the areas around Middens 3, 10, 5, 7, 8, and 11 had to be ignored. Further the samples from Middens 3, 8 and 10 were also too small. Samples from Areas 1, 6 and 9 and Midden 11 were also small but have been included in the discussion. Figure 11 shows the defined cultural areas. The data giving rise to these observations can be found in Appendix C: Tables C-19 through C-23.
Chert Type Observations In considering the distribution of burned or unidentifiable chert, the percentage of burned chert on the site does vary between areas. In looking at the midden areas, the amount of burned chert varies from 3.7 % to 43%, a much wider variation than that observed for the Brian Site. The statistically significant difference is as follows.
1. Middens 1 and 2 have a higher percentage of burned chert than Middens 4, 5 and 6.
Analysis of the distribution of the various chert types against the cultural areas yields the following observations.
2. Area 6 and Middens 5, 6 and 7 have low values for Kettle Point chert while Middens 1, 2, 4 and 9 have percentages of Kettle Point chert that are higher with statistical significance.
3. Areas 2 and 4 have higher percentage of Kettle Point chert than Middens 6 and 7 with statistical significance.
4. Areas 1 and 9 have high percentages of Kettle Point chert but the samples from both areas are small.
5. Middens 5, 6 and 7 and Area 6 have higher percentages of Onondaga chert than Middens 1, 2, 4 and 9 and Area 2 with statistical significance.
6. Areas 1 and 9 have no Onondaga chert but the samples are small.
7. Area 4 is an anomaly. It ranked in the group of areas that have high Kettle Point chert but it also has a high percentage of Onondaga chert. It has a higher percentage of Onondaga than Middens 1, 2, 4,and 9 and Area 2 with statistical significance.
8. Area 4 and Midden 7 have lower percentages of local till chert than many other areas with statistical significance.
Flake Type Analysis An analysis was run looking at the distribution of the various flake types against the cultural areas with the following result.
9. There are no statistically significant differences across any of the cultural areas.
General Comments In considering the observations on the Drumholm site, it is evident that differences between the same groups of areas can be found in both the analysis of the distribution of Kettle Point chert and Onondaga chert. Area 6 and Middens 5, 6 and 7 are different from Middens 1, 2, 4, and 9. In addition, Areas 1, 2 and 9 are high in Kettle Point and low in Onondaga. In looking at these trends as a whole, it became obvious that the areas with high Kettle Point and low Onondaga chert were to the north while the areas with the reverse were to the south. Consequently, a new definition of cultural areas that split the site into two halves, north and south, was developed. Middens 1, 2, 3, 10, and 11 and their associated areas are in the north unit. Middens 5, 7 and 8 and their associated areas are in the south unit. Middens 4, 6 and 9 and their associated areas have been divided between the two units. The boundary between these two units is marked on Figure 11. All of the following differences are statistically significant.
10. When the spatial analysis was run the following confidence intervals result.
$ Kettle Point chert was 38.7 +/- 8.2% in the north and 9.1 +/- 5.7% in the south half.
$ Onondaga chert was 19 +/- 6.6% in the north and 64.3 +/- 9.4% in the south half.
$ Local Till chert was 40.9 +/- 8.2% in the north and 25.3 +/- 8.6% in the south half.
Clearly the north half of the site had more use of Kettle Point chert than did the south half and the south half had better access to Onondaga chert than did the north half.
Again it appears that a shortfall in access to high quality chert was compensated for with greater use of local till chert in both areas of the site.
One area of the site that does not fit this generalization well is the area around Midden 4. This area is high in both Onondaga and Kettle Point chert (and consequently very low for local till chert). This area may represent a special zone within the site that was used preferentially for high quality chert working but it should also be noted that this area was split to create the north/south site division just described. In fact, Onondaga chert tends occur in the south half of the area while Kettle Point tends to be found more in the north half. However, neither of these distributions is statistically significant. While it might be unwise to assign any special significance to Area 4 without more evidence, it should be noted that a similar area was found at the Brian site above and it was near the centre of the site.
Chapter 6: Inter-Site Spatial Analysis This chapter provides an inter-site comparison of the data regarding lithic acquisition across the analyzed sites. Appendix F includes a one page summary of the basic lithic data for each site. In this Appendix, each site has three tables. The first table is the raw counts of the breakdown of the eight flake types by the five chert types. During the analysis of the debitage, it quickly became evident that very few unifacial retouch flakes or bipolar flakes were being identified. Consequently, the second table shows the percentages of the five predominant flake types against the five chert types. The third table breaks down the various formal artefact types against the five chert types and calculates some summary percentages. Given the desire to include more sites that are represented by only surface collections or smaller samples, the number of artefacts analyzed is frequently very low, making detailed analysis problematic. The other data used in the analysis concerns the specific sites. Appendix A lists site specific information including the cultural period and the type of site. Data regarding the distance from the respective sources and the percentage of each chert type used in the regression analysis is found in Appendix D, Table D-1. The location of the site was also carried as part of the original spreadsheet in this table, but Ministry of Culture regulations do not permit disclosure of this information so it has been intentionally left out of the Appendix. As in the previous chapter the percentage of any given chert source type for a given site is calculated by dividing the number of flakes of that type against the total of all flakes of identifiable chert types exclusive of unidentified chert (UID in the tables). All flakes including unidentified are included when comparisons are being done with respect to the morphology.
The following analysis is broken down into several discrete steps. First, for the debitage, the distance decay of the two imported chert types, Kettle Point and Onondaga, is examined. Second, morphological variation both in total and by chert type is considered. Third, a comparison is made of Neutral villages versus Cabin sites. Fourth, variation in formal and informal artefacts is examined. And finally, five individual sites are examined that vary significantly from the norm.
Kettle Point Chert Distance Decay The analysis of the use of Kettle Point chert involved creating Table D-1 as shown in Appendix D that gives the percentage of Kettle Point chert and the distance from the source. The percentage is drawn from the data in Appendix F and the distance was calculated from the UTM grid reference for each site to the UTM grid reference for Kettle Point (4,785,300 north, 417,400 east). Next, the information was plotted as a scatter plot with the distance as the X-axis and the percentage as the Y-axis. Each of the three periods was plotted as a distinct set of points so that changes over time would be apparent. Also for reasons discussed later, the LOI sites are broken down by villages and cabin sites. Finally, regression lines were run on each of the four groups of sites. The results are shown in Figures 12 through 16. In adding the regression lines four different equation forms were tested using Microsoft Excel - 97. These were linear, logarithmic, power and exponential. Each line was added and the parameters of the resulting equation determined and the R-Squared value calculated determining goodness of fit.
This statistic (Wonnacott and Wonnacott 1990:486) varies from 0 to 1 with higher numbers representing a better fit to the data. A value of zero would represent absolutely no correlation between the line and the data and a value of 1 would be a perfect fit.
While residuals could have been compared, it was felt that, as the samples used were not large, this would be extending the analysis beyond what the data could reasonably support.
The first general observation that can be made is that the site locations are not absolutely ideal in that they do not occur evenly over the entire study area. The Glen Meyer sites have perhaps the widest distribution occurring across the entire area. Neutral sites are well-distributed but do not occur in the westerly portion of the area. The MOI sites occur across the area but when broken down by Uren versus Middleport substage, there are not many Uren sites and all of the Middleport sites tend to occur in a narrow belt in the centre of the area causing some difficulty for the analysis. There are Middleport substage sites in the Whittaker Lake cluster at the eastern edge of the study area, for example, the Messenger Site (AfHf-3). However, there are no reasonable samples from this site as it has never had any archaeological work done on it. There may also be Middleport substage sites in the area between London and Dorchester. However, samples from this area are meager and at least one large village site in the Waubuno area has been destroyed by gravel pit operations (Keron 1986). There may be additional sites in this area but it has received little survey attention. Certainly, the addition of more easterly Middleport substage sites would greatly enhance this analysis.
Considering all the sites as a whole, the plot of the various percentages against distance, at first, seemed to have no evident patterning. However, once the sites were separated out by time period, it became evident there were very significant changes from one period to the next that obscured each other when lumped together.
First, while there appeared to be a very obvious linear decay to the Glen Meyer sites, the interpolated line was flat showing no decay with space. It quickly became evident that this was caused by three outliers; sites that had percentages of chert source types that were widely different from the obvious norm (see Figure 12). After careful consideration described below, these were excluded and the regression line produced a high R-Squared value.
Upon completion of analysis described later in this chapter, it became evident that the Neutral sites would need to be broken down by villages and cabin sites for the distance decay analysis. This improved the fit of the regression line for the villages but the cabin sites were more problematic as the iteration that included all of the cabin sites produced very low R-squared values. Next, a second iteration was attempted excluding two outliers. This produced better R-Squared values but, unlike the reasoning used to exclude the Glen Myer outliers, there is no evident reason for excluding these two. In fact, results of other analysis described below indicate that consistency is not to be expected in cabin sites. In all cases, four different trend-lines were examined for goodness of fit, measured by the R-Squared statistic. The equations forms were linear, logarithmic, exponential and power. The results of this analysis are presented in Appendix D, Table D-2. The following observations can be made.
1. Glen Meyer use of Kettle Point chert starts at around 40% in the west and declines gradually to 30% in the east. A number of sites also tend to create a very obvious line of slowly declining chert usage (Figure 12).
2. There are three Glen Meyer sites that are greatly divergent from this line. One, the McGrath Site, has nearly 100% Kettle Point chert in the debitage and is interpreted as a temporary camp site. The two other divergent sites with almost no Kettle Point chert are cabin sites along the south edge of the cluster of sites in the Caradoc sand plain (Figure 12).
3. Use of Kettle Point chert drops sharply in the Uren Substage being consistently under 10%, suggesting restricted access to the source. See the Uren sites in Table D-1 in Appendix D.
4. During the Middleport substage there seems to be increased access to Kettle Point chert although most sites are still restricted in comparison to Glen Meyer (Figure 13).
5. During Neutral times, looking at the villages, access is not only restored but it seems to become the preferred chert source especially in the western sites in the London area. For example the Lawson site debitage at 85% is almost exclusively Kettle Point chert (Figure 14).
6. Neutral sites in the Lambeth area, while at a similar distance from the source as the Lawson Site, have lower percentages generally in the 50-60% range (Figure 14).
7. Neutral sites in the Pond Mills area have percentages in the 20-30% range (Figure 14).
8. Neutral sites in the Whittaker Lake area have the lowest percentage usually around 10% Kettle Point chert (Figure 14).
9. Looking at the Neutral cabin sites, the two trend lines presented both lie to the left of the trend line of the Neutral villages indicating a lower use of Kettle Point chert (Figure 15).
10. The Glen Meyer trend line is best approximated by a linear equation (Figure 12, Table D-2). The R-Squared value is high at.809.
11. The MOI trend line is best approximated by a power equation. The R-Squared values are low for all equation types indicating a lot of variability that is not directly related to distance (Figure 13, Table D-2).