Introduction
   
        Experimental knapping has considerable potential to teach us about past behaviors. Lithic tool production is especially amenable to experimental research because the raw material is not plastic and the process is reductive; there are limited ways in which the reduction sequence can proceed.  Given the nature of lithic reduction, it is also possible to learn something about the prehistoric production process even when it is not imitated exactly.  Because of the limitations imposed by the raw material, differences in end results can be instructive.
       
        The purpose of this project is to learn about the process of Hopewell blade production through imperfect replication of the artifacts by a novice knapper (KCN).  We will present some preliminary results of a project very much in progress.  So far, nodules of two different raw material types have been reduced producing over 150 complete blades and over 250 blade fragments.  The blades were produced in two separate production events by a single knapper using direct soft hammer percussion (Figure 1).  The total time of production, though not strictly timed, was less than an hour.  This observation in itself has implications for archaeological interpretations of blade production sites.
 
Image
Figure 1: A Novice?s Toolkit. The tools used in the production of the experimental blade collection: 1 large moose antler billet, 1 abrading stone, and leather gloves and leg pad for protection.
 
 
        A total of 414 blades and fragments have been partially analyzed.  The attributes recorded in this analysis are modeled directly after Nolan (2005) (see also Greber et al. 1981 and Ruby 1997).  The continuous attributes we will discuss are blade length (N=164), width (N=399; Figure 2), thickness below the bulb of force (TbBOF; N=235), maximum thickness (N=412), platform width (N=13), and platform thickness (N=13).  Only 334 blades have had their discrete attributes recorded, but this is still a fairly large sample to work with.  The categorical attributes are completeness (complete, proximal, medial, or distal), cortex presence, facet count, cross-sectional shape (Figure 3), and termination type (feather, hinge, overpass).
 
Image
Figure 2: An Illustration of Variation. This figure portrays the range of variation present in the width of the experimental blades. The left column contains blades whose widths are 1 standard deviation or more below the mean. The middle column blades have widths that approximate the mean. The right column of blades has widths 2 standard deviations or more above the mean.
 
Image
Figure 3: Trends In Size and Shape. This figure gives our best approximation of the trends in removal size and shape as the removal sequence proceeded for each nodule. The sequences for each nodule go from top to bottom, left to right. The top two rows are from the Wyandotte nodule, the bottom two from the Burlington. Notice the large disc-shaped, cortex covered flake in the upper left-hand corner. This was the first removal from the Wyandotte half nodule.
 
        In the rest of this paper, we will present summaries for the attributes of the experimental collection and comparisons with the same attributes for the Turner Workshop blades (Nolan 2005; Nolan, Seeman, and Theler 2006).  The Turner Workshop is located in Hamilton County, Ohio on a terrace south of the Little Miami River within 300 meters of the Turner Earthworks (33HA26) and Turner Parallel walls (33HA263) (Nolan 2005:Figure 1).  The collection is the result of intensive and extensive surface collection over roughly one acre of land by Clyde and Jim Theler in the 1960s and 1970s.  For the metric attributes, we compare both central tendency and relative variability.  We use the Coefficient of Variation as a measure of relative variability.  Formulas for statistical comparisons of CVs were taken from Sokal and Braumann (1980:61-62). 
 
Results
 Continuous Attributes
 
        There is typically a wide range of variation in length in Hopewell blade collections. This is consistent with Ruby’s (1997:235) conclusion that little control was exercised over this attribute.  The experimental mean is 32.8 mm, with a standard deviation of 11.48 (Figure 4).  As illustrated in Figure 5, the distribution for width is slightly skewed to the right.  The Turner Workshop histogram for this attribute is similarly skewed (Nolan 2005:Figure 3).  The mean value for this attribute is 15.88 mm (s=6.93).  This mean may decrease after we measure the pile of small debitage left in the lab, resulting in a slightly more normal distribution.  The TbBOF in the experimental collection has a mean of 2.6 mm with a standard deviation of 1.5.  Like width, it has a distribution skewed slightly to the right (Figure 6).  The experimental collection yielded a mean of 2.94 mm (s=1.69) for maximum thickness with a distribution similar to that of the TbBOF (Figure 7). Like width, both thickness measure distributions are similar in shape to those observed for the Turner Workshop collection (Nolan 2005:Figures 4, 5).
 
Image
Figure 4: Experimental Blade Length Histogram.
 
Image
Figure 5: Experimental Blade Width Histogram.
 
 
Image
Figure 6: Experimental Blade TbBOF Histogram.
 
Image
Figure 7: Experimental Blade Maximum Thickness Histogram.
 

We have not had time to record all of the platforms for the experimental blades.  However, we did take a random sample of 20 blades to measure their platform attributes for this paper.  Of those 20, only 13 had intact platforms.  The mean for lateral platform width is 8.42 mm with a standard deviation of 4.58 (min 1.6mm; max 18.1 mm; Table 1).  Dorso-ventral platform thickness averages 3.47 mm with a standard deviation of 2.05 (min .8 mm; max 7.8 mm; Table 1).  We refrain from statistical comparisons due to the sample size but we will say that these values are much larger than those recorded for the Turner Workshop collection (PW: 4.61 mm; PT: 1.63 mm; Nolan 2005:Table 1). 

 

Table 1: Experimental Platform Attributes.  This table presents our measurements of a random sample of 20 experimental blade platforms.

 

PW

PT

N

Valid

13

13

 

Missing

7

7

Mean

8.423

3.469

Median

8.700

3.400

Std. Deviation

4.5844

2.0524

Minimum

1.6

.8

Maximum

18.1

7.8

 
        Table 2 shows the statistical comparisons of means between the experimental collection and the Turner Workshop collection.  Our length was greater than that of the Turner Workshop, but not significantly so.  Our width, however, is quite significantly greater than the width of the Turner collection.  In fact, our mean value is almost 4 mm greater.  That value may decrease somewhat after we include the debitage pile in the sample, but not enough to make this statistical difference disappear.  The thickness below the bulb of force in the experimental collection is less than that of the Turner Workshop collection, but not significantly so.  Our maximum thickness is less than the Turner thickness and the difference is significant. 

Table 2: Central Tendency Comparison.  This table presents the results of comparisons of central tendency between our experimental blades (EN) and the Turner Workshop (TW) collection (Nolan 2005; Nolan, Seeman, and Theler 2006).

 

 Length

Width

TbBOF

Maximun Thickness

EN

  s

32.8

11.48

15.88

6.93

2.6

1.5

2.94

1.69

TW

  s

31.17

8.68

11.96

4.11

2.71

1.2

3.25

1.44

  t

1.958

16.563

-1.369

-4.137

p

.051

<.0001

.171

<.0001

 

Table 3 presents the results of the relative variability comparisons.  For all the attributes compared, the experimental blades exhibit a wider range of variation, that is, larger coefficients of variation, than the Turner Workshop blades.  With the exception of length, these differences are very highly significant.

 

Table 3: Relative Variability Comparison.  This table presents the comparisons of Coefficients of Variation between our experimental blades (EN) and the Turner Workshop (TW) blades.  Formulas for comparisons of Coefficients of Variation were taken from Sokal and Braumann (1980).

 

 Length

Width

TbBOF

Maximum Thickness

EN

  s

34.41

1.9

43.67

1.82

57.58

3.42

57.52

2.58

TW

  s

27.85

.87

34.34

.45

44.22

.88

44.21

.61

  t

-3.1349

-4.985

-3.7768

-5.0156

p <

.003

.0001

.0001

.0001

 

Categorical Attributes.

        With respect to completeness (Figure 8), it is intriguing to notice the different proportions of especially complete blades and medial fragments.  The high percentage of complete blades is because our blades were not plowed under for hundreds of years.  The frequency of medial fragments relative to proximal and distal fragments is puzzling.  Since each blade can produce only one proximal and one distal fragment but potentially many medial fragments, there could be many more medial fragments than proximal or distal.  This is frequently the case for archaeological blade assemblages (e.g. Genheimer 1996).  We assume one of two scenarios: either our pile of debitage contains many of our missing medial fragments, or, since our collection has not been subject to plowing and destruction, our blades are more likely to have broken only once leaving mostly proximal and distal fragments.  In our graph of cortex presence (Figure 9), we clearly have a higher percentage of blades with cortex present than the Turner Workshop collection.  Our collection could be biased by the fact that the Wyandotte blade-core was completely covered in cortex.  It seems likely that the Hopewell knappers were getting nodules that had been previously decorticated.

Image
Figure 8: Blade Completeness Frequency Comparison.

Image
Figure 9: Blade Cortex Presence/Absence Comparison.
 

        Pacheco (1997) suggested that Hopewell blade producers desired three facets and a trapezoidal cross-section.  Indeed, the majority of the Turner Workshop blades conform to this expectation.  In contrast, two facets and a triangular cross-section predominate in the experimental collection (Figures 10 and 11).  It is generally assumed that blade producers were seeking the neat edge of a feather termination.  If this is indeed the case, we can note that we had greater success in producing a feather termination than the Hopewell blade producers (Figure 12).  It appears that the Hopewell blade producers exercised little control over termination type.  This is consistent with the conclusion that the Hopewell blade industry was not specialized (Nolan, Seeman, and Theler 2006). 

 
 
Image
Figure 10: Blade Facet Count Comparison.
 
Image
Figure 11: Blade Cross-Section Comparison.
 

Lessons Learned
Blade-core size/face curvature. 

        One of the more interesting differences between the experiments and the Turner blades is the differences in width.  We appear to have made “Illinois-like” blades in our experiments.  Several factors contribute to the realized width of a blade.  Among these are removal face topography and placement of force.  Both of these likely contributed to the differences here.  The extreme thickness of several of the measured platforms in the experimental sample indicates application of force away from the platform removal face intersection.  However, we think it is likely that the size of the nodules used in the experiments and curvature of the chosen removal faces contributed the most to the differences in width (Figure 13).  Greber et al. (1981:Tables 9 and 13) report that the blade-cores from the Illinois Hopewell Snyders site were on average thicker and wider than the Liberty blade-cores.  It may be that differences observed between Illinois and Ohio Hopewell blade widths can be attributed to the use of different sized and shaped nodules as blade-core preforms.

 

Image
Figure 12: Blade Termination Type Comparison.
 

Removal Technique. 

        One of the postulated differences between Ohio and Illinois blade industries has been the method of blade removal.  The smaller platforms of Ohio-affiliated sites have been attributed to indirect percussion (Greber et al. 1981).  While we have limited data on platform dimensions, the blades produced in this experiment have platforms that are (with a single exception; Table 1) within the range of the Turner Workshop platforms (Nolan 2005:Table 1).  While it should not be overlooked that the means for both platform attributes are extremely different, these experiments demonstrate that it is possible to produce small platforms with direct percussion.  Small platforms can be achieved by grazing the platform with the percussor.

Formation Processes. 

        A very interesting lesson to take away from these experiments is the fact that blade fragments were created during production and not by any use or post depositional disturbances such as plowing.  We are not aware of a way to sort out blade fragments created by use, post depositional factors, or simply as a byproduct of production.  However, this should be kept in mind when considering the history of assemblage formation.

“Prepared” blade-core. 

        Several discussions of blade production mention the presence of a prepared core as a defining characteristic of a blade industry.  A prepared core is described as a nodule that has a platform surface that has been prepared and having been shaped to a ridge on one surface to guide the initial removal (Odell 1994; Yerkes 1994).  However, preparation of the nodule is not a prerequisite to blade production.  Neither of the nodules used in these experiments were prepared in the manner just described.  The Burlington nodule was tabular with a naturally flat surface and a pre-existing ridge on one face.  The Wyandotte nodule was prepared slightly more; the nodule was purchased as a sawed half-nodule, thus creating a “prepared” platform surface.  Still, no steps were taken to form a ridge on the removal face.  Production was initiated simply by removing a large disc-shaped flake (Figure 3).  The two irregular arris scars thus created were the basis for all subsequent removals.  It may be that the difficulty and requisite effort for Hopewell blade production has consistently been overestimated.

Skill. 

        Two separate proxy measures of knapper skill were used when analyzing the Turner Workshop sample: frequency of non-feather terminations and relative variability (Nolan 2005; Nolan Seeman and Theler 2006).  In this case, these two measures give seemingly contradictory results.  Error rates, as measured by termination type, for the experimental sample are lower than those recorded for the Turner Workshop (34.8%, and 43%, respectively; Figure 12).  On the other hand, the experimental blades exhibit a much wider range of variation than the Turner blades.  All the attributes compared, with the exception of length, showed significant differences (α=.001) in coefficients of variation.  This indicates the prehistoric blade producers were able to approximate an ideal size and shape more consistently.  Another form of knapping error may account for the extreme variability in the experimental blades.  In the platform sizes observed thus far, the position of the application of force in the experiments was highly variable and often at a distance from the platform edge (Table 1).  This contributed significantly to variation in width and thickness. 

Effort and Blades per blade-core. 

        We will discuss two more related lessons: the amount of time investment blade production requires and the number of blades that can come off a single blade-core.  As noted in the introduction, the whole of the experimental collection, approximately 230 separate removals (complete blades plus proximal fragments), was created by a single individual in less than an hour.  Each knapping session lasted approximately 20-30 minutes.  The take home message is that very large numbers of this type of artifact can be created, and likely were created, with very little time and effort.  Additionally, the average number of blades per blade-core in this experiment is 115.  That is, it took one rock, one antler, one person, and 30 minutes to produce 115 multipurpose stone tools.

        There are other potential insights that can be gleaned from the data thus far collected; however, those discussed here should suffice for now to demonstrate the usefulness of this type of project.  Of course, all of these interpretations are tentative and subject to revision as experimentation and analysis continue.

 

Future Directions
Sorting and Recording. 

        There are still two piles of debris, one from each nodule, that remain unsorted.  The fragments in those piles will supplement the data presented here and likely fill in some of the “gaps” in our histograms and bar graphs.  Given that these debris are tiny fragments, our means for some attributes may decrease.  Obviously, we must record the attributes of those fragments yet to be sorted, but we must also finish recording the categorical attributes of the blade fragments.  We are especially anxious to record the platform attributes (size and exterior angle) for the entire experimental sample. 

Additional Experiments. 

        As mentioned in the introduction, this is a project very much in progress; a project in its infancy even.  While there are many possible directions this project could take, we have definite plans for the near future.  We want to increase our sample size with new nodules and new raw material types.   We currently have at least 3 potential blade-cores: 1 Upper Mercer, 1 heat-treated Flint Ridge, and 1 raw Flint Ridge. 

        We would also like to incorporate additional blade producers into the experiment.  Of particular interest would be someone of a higher skill level and potentially multiple knappers and a range of skill levels.  With more producers we could integrate new removal techniques.  With these additional experiments some of the tentative interpretations proposed here could be tested. 

Refitting. 

        Another very fruitful direction for this project would be to undertake the task of refitting the blades and fragments back to the blade-core to reconstruct each specific production sequence.  This analysis could draw out more detailed similarities and contrasts between the prehistoric and experimental processes. 

        In summary, experimental knapping can aid in the reconstruction of past patterns of behavior.  Specifically here we have shed light on some aspects of the Hopewell blade productions process through comparison of our experimentally produced blades and those from the Turner Workshop.  This project and others like it have much to tell us about prehistoric organization of production.

 

References Cited

Genheimer, Robert A.

1996    Bladelets are Tools too: The Predominance of Bladelets Among Formal Tools at Ohio Hopewell Sites. In A View From the Core, edited by P. J. Pacheco, pp.92-107. Ohio Archaeological Council, Columbus.

Greber, N’omi, Richards S. Davis, and Ann S. Dufresne

1981    The Micro Component of the Ohio Hopewell Lithic Technology: Bladelets.  In The Research Potential of Anthropological Museum Collections, edited by A. E. Cantwell, J. B. Griffin, and N. A. Rothschild, pp. 489-528.  Annals of the New York Academy of Sciences 376.  New York.

Nolan, Kevin C.

2005    The Ohio Hopewell Blade Industry and Craft Specialization: A Comparative Analysis.  Unpublished MA Thesis, Kent State University.

Nolan, Kevin C., Mark F. Seeman, James L. Theler

2006    Hopewell Blade Production at the Turner Workshop. Paper presented at the 71st SAA.

Odell, George H.

1994    The Role of Stone Bladelets in Middle Woodland Society. American Antiquity 59: 102-120.

Pacheco, Paul J.

1997    Ohio Middle Woodland Intracommunity Settlement Variability: A Case Study from the Licking Valley.  In Ohio Hopewell Community Organization, edited by W. S. Dancey and P. J. Pacheco, pp. 41-84.  The Kent State University Press, Kent, Ohio

Ruby, Bret J.

1997    The Mann Phase Hopewellian Subsistence and Settlement Adaptations in the Wabash Lowlands of Southwest Indiana.  PhD. Dissertation, Indiana University.

Sokal, Robert R., and Carlos A. Braumann

1980        Significance Tests for Coefficients of Variation and Variability Profiles.  Systematic Zoology 29:50-66.

Yerkes, Richard W.

1994    A Consideration of the Function of Ohio Hopewell Bladelets.  Lithic Technology 19(2): 109-127.