SUMMARY AND CONCLUSIONS
Patricia E. Miller
The studies described in this report involved the evaluation of existing archaeological site data in three watersheds identified by the BHP's survey priorities policy as having a low priority for additional site surveys. The driving assumptions of the survey priorities policy are that most upland sites are multicomponent lithic scatters that only provide information on settlement patterns; that in the 19 low-priority watersheds there are already enough of these sites to interpret settlement patterns; and that as a result additional sites would not provide significant new information and therefore would not be eligible for the National Register of Historic Places. Since legal mandates require protection only for National Register-eligible sites, additional survey is not requested by the BHP in upland areas of the 19 watersheds, except in areas with a high probability for site types other than multicomponent lithic scatters. The goals of PAC's settlement pattern research priorities project were to evaluate the policy's assumptions by determining whether settlement patterns and other important research issues can be addressed with the existing data in the studied watersheds; considering whether the identification and study of additional sites would be likely to provide needed information; and providing recommendations on where such sites would be located. The results of the project are discussed below in terms of these goals.
Are most upland sites multicomponent lithic scatters without features?
Examination of the recorded site data from the Conodoguinet drainage revealed 16 of the 37 dated sites were multicomponent and 21 were single component (Table 1). In the Brandywine watershed, 33 sites were single component and 49 were multicomponent. In the Conemaugh-Blacklick watershed 47 were single component and 76 sites were multicomponent. In the Meyersdale survey summarized by Chiarulli, eight of the 17 identified sites were single component. However, some of the single component sites in the PASS file data may actually have other components that were not identified in Phase I survey or surface collections. Of the upland sites with Phase II surveys in the Brandywine watershed, four were single component, six were multicomponent, and eleven produced no temporal information. Of the three upland sites in the Conodoguinet watershed where Phase II surveys were performed, one site was clearly multicomponent, one site was attributed to both the Late Archaic and Late Woodland on the basis of triangular points, and one site was attributed to the Transitional Period on the basis of rhyolite debitage. In the Conemaugh-Blacklick watershed, however, all six of the Phase II sites were multicomponent. Overall, the site data from the three watersheds indicate that while undated and multicomponent sites predominate, single component sites are well represented in the database.
Table 1: Single Component vs. Multicomponent
According to PASS file data for all of Pennsylvania, only 5% of sites in non-riverine settings have features (BHP 1996:7). However, only sites that have undergone Phase II survey or other detailed field investigations are likely to have had subsurface testing sufficient to identify features. In the Brandywine Creek watershed, information was available to the researchers on 19 upland sites that have been evaluated through Phase II surveys, four (21%) of which had features. In the Conodoguinet watershed, only three upland sites have had Phase II surveys, none of which were found to have features. Of six sites in the Conemaugh-Blacklick watershed where Phase II investigations were conducted, four (66%) had features.
Chiarulli summarized excavations of upland sites in the Conemaugh-Blacklick watershed and surrounding areas. Many of these sites produced chronologically diagnostic artifacts and datable features, although most of the sites were multicomponent. Fourteen upland sites found during the Meyersdale project were investigated through Phase II survey and three (21%) were found to contain features. All three sites, as well as a rock shelter with no features, were determined eligible for the National Register.
Thus, the data indicate that features are more prevalent on upland sites than the overall PASS file data suggests. In addition, although the majority of dated sites in the three watersheds are multicomponent, single component sites comprise as much as 45% of the dated sites. Although some of the single component sites may have other components that have not been identified, it is notable that a substantial proportion of sites with Phase II level investigations are found to contain only one prehistoric component.
Do the three watersheds have enough recorded sites to address the issue of settlement patterns?
Settlement pattern studies at their most basic level focus on identifying consistencies in site locations and how patterns of site location vary over time. However, in behavioral terms settlement patterns represent strategies of land and resource use that cannot be interpreted on the basis of site distributions alone. Settlement strategies can only be completely understood if information on site function, seasonality, and duration of occupation is available. Tools from datable contexts, through typological and use wear analysis, can provide information on the types of activities performed at a site, which is indicative of site function. Feature analysis often provides information on site function and seasonality. The confirmed absence of features such as storage pits, postmolds, and elaborate hearths is also evidence of site function, suggesting short-term or special-purpose use. The diversity of the tool inventory can be related to both site function and the duration of occupation.
Each of the three watershed studies examined the existing PASS file data provided by the BHP in electronic format to determine whether settlement pattern research issues can be addressed without additional data. In the Conodoguinet watershed, after eliminating three historic and two erroneously recorded sites, only 70 upland sites were in the database. Of these, only 37 upland sites had chronological data and most of these sites dated to the Late Archaic and Late Woodland Periods. Other periods were represented by less than 10 sites each. Slightly more upland sites were recorded in the Conemaugh-Blacklick watershed-138 prehistoric sites. Of the 138 sites, 123 contained one or more datable components. Still, Paleoindian and Early Archaic sites were represented by fewer than 10 components each and the Middle Archaic, Early Woodland, and Middle Woodland periods were represented by only 22, 15, and 15 components respectively. In the Brandywine watershed 218 sites were present in upland settings, 90 of which contained dated components. All of the periods that were poorly represented in the other two watersheds were represented by fewer than 10 components in the Brandywine watershed. Thus, the number of known sites is limited overall, but especially for those periods of occupation that we know least about.
Table 2: Number of Datable Components*
|Archaic||Transitional/ Terminal Archaic||Woodland|
*Excludes undefined Archaic and Woodland components
All of the researchers in this project noted problems and biases in the site data. Many of the sites do not have information for all of the environmental and archaeological variables that would be needed for settlement pattern research. Many have been recorded by amateur archaeologists and have not been field checked. Errors such as duplicate site numbers have also been identified. Collector bias toward plowed fields and riverine settings is well known, so upland sites are almost certainly under represented in the site data. This bias is of concern in evaluating the relative intensity of prehistoric use of upland vs. riverine areas. Some of the problems can be remedied by careful checking of the site data. Dr. Dean Snow, in his review of the reports for this project, noted that a formal program to field check the PASS file data is needed. Until such a program can be designed and implemented, any researcher using PASS file sites to develop predictive models or interpretations of settlement patterns should field verify the accuracy of the site data they use.
As noted above, settlement pattern research involves more than being able to predict where sites are located. To understand settlement strategies and how such strategies changed over time, information on site function is needed. Most sites in the PASS file data are listed as open habitations or lithic scatters, which provides no information on site function or related activities. In cultural resource management studies, sites with few artifacts and no features are generally interpreted as short-term or special-purpose occupations as opposed to base camps, hamlets, or villages. This is a rudimentary classification but information for more specific and accurate assignment is often not available on upland sites. On multicomponent sites tools generally cannot be associated with specific chronological periods, making site function difficult to determine. The number and types of features may indicate site function and duration of occupation, with storage pits and house patterns indicating habitation sites as opposed to shallow hearths that indicate shorter-term occupation. But Phase II or comparable testing is necessary to confirm the presence or absence of features and to provide information on feature types that relate to site function. Phase II testing also increases the chances that all components present on a site will be identified.
Unfortunately, few Phase II surveys have been performed on upland sites in these watersheds. There are three in the Conodoguinet watershed, six in the Conemaugh-Blacklick watershed, and 19 in the Brandywine Creek watershed. While it could be argued that existing sites in these watersheds are protected and more intensive investigations could be performed in the future, it is also true that the sites are not protected from private actions and could, therefore, be destroyed in the future.
Settlement pattern research also benefits from surveys that do not produce sites by identifying environmental settings that were not utilized prehistorically. There is clearly a political problem in maintaining support for surveys that do not produce sites, since neither permittees, agencies, nor the public would readily recognize the benefits of such surveys. However, the Geographic Information System (GIS) has tremendous potential as a tool in settlement pattern research using data from systematic site surveys (Duncan and Beckman 2000; Kvamme 1990). Because of the flexibility and data management capacities of GIS, environmental settings of sites can be compared with surveyed settings that have no sites. However, statistically reliable models require data from systematic surveys to eliminate the biases in coverage discussed above. But systematic survey in the three watersheds is extremely limited. In the Brandywine watershed, systematic surveys covered only 1976.71 acres or 1.02% of the watershed. In the Conodoguinet watershed, only 19 Phase I surveys have been performed, covering an area of 2.3 km2 or 0.15% of the watershed. In the Conemaugh-Blacklick watershed less than 1% of the area has been surveyed.
Thus, the existing site data in the three watersheds have been shown to contain an insufficient number of dated components, especially from periods other than the Late Archaic and Late Woodland, as well as an insufficient number components for which site function, seasonality, and duration of occupation is known, to interpret prehistoric settlement patterns. The data also indicate that sites with features, which would provide information on site function, are more prevalent than suggested by the statewide data base. Likewise, single component sites do occur with some frequency in upland settings, though less often than multicomponent sites. These conclusions indicate that, given the small proportion of the watersheds that have been surveyed to date, the additional data needed to address the issue of settlement patterns could be recovered through additional site survey and excavation.
What other research issues could be addressed by sites in upland settings? What information defines a significant site?
Research issues for the three watersheds have been discussed in the preceding chapters of this report. General categories of research issues other than settlement patterns can be identified that are of interest in research on prehistoric behavior in non-riverine settings. The list below is not exhaustive, nor can it account for research issues that will be defined in the future as new information and analytic techniques become available. However, in the context of this project an important question is whether upland sites are likely to contain the data needed to address issues that are currently considered important in the study of prehistoric behavior.
Material Culture - The identification of material culture and how it varies over time and space is basic to an understanding of cultural boundaries and technological change. Defining material culture involves the delineation of regional artifact typologies, as well as hafted biface and ceramic chronologies. Material culture provides information on other aspects of prehistoric behavior such as degree of mobility, subsistence activities, and technology. To address issues related to material culture, a site would need to contain artifacts in datable contexts.
Lithic Procurement and Use - A number of research questions are subsumed under this general topic. How are lithic reduction activities patterned around quarry locations? How does distance from a lithic source affect its frequency of use? How does lithic resource use vary over time? To what degree are Paleoindian and Early Archaic populations tethered to high-quality lithic materials. Upland sites are extremely important in this area of research since most source locations are in upland settings. Locating quarry sites is a requirement for addressing most of these research issues. In addition, the material types of chronologically diagnostic tools can provide information on how lithic resource use varies over time and with quarry distance. The proportion specific material types in debitage assemblages is also useful if the assemblage can be dated.
Subsistence - Small upland sites with low artifact densities are often interpreted as special-purpose procurement sites that may be related to seasonal subsistence tasks. However, these interpretations are often based on the absence of data suggesting longer-term base camp occupations. Models involving base camps in riverine settings and satellite specialized sites in the uplands require additional support. Data that could occur on upland sites and would address this research issue include residues on chronologically diagnostic lithic tools, ceramics, or steatite; botanical and faunal information from datable features; and tools related to food procurement and preparation, such as nutting stones, mortars/pestles, hoes, and netsinkers, from datable contexts.
Trade and Exchange - Trade and exchange are considered to be important strategies of adaptation in that they establish and reinforce social networks that are useful in times of resource stress. Interaction spheres can be identified by the types of exotic lithic materials found in an assemblage and by the distribution of those materials across the landscape. Evidence suggests that the importance of trade and exchange, as well as the boundaries of interaction spheres varied over time. Therefore, only artifacts on single-component sites or in other datable contexts could contribute to this research issue.
Social organization - The social organization of prehistoric groups occupying upland sites is difficult to determine on the basis of upland site data. In the Ohio River valley of Pennsylvania, where villages or hamlets often occur on ridge tops and saddles, the intrasite patterning of houses and features, as well as the presence or absence of artifacts related to status differentiation or ceremonial activities, could provide information on social organization. In general, aspects of social organization can be revealed by patterns of sites of different function across the landscape. Base camps of different sizes and lengths of occupation can be interpreted as reflecting seasonal patterns of fissioning and aggregation. Information on site function and periods and lengths of site occupation are needed.
Most investigators recognize that site types other than multicomponent lithic scatters are present in upland settings. Such sites include Late Prehistoric villages and hamlets, single-component base camps, quarry sites and related processing stations, and rock shelters. The high probability upland settings requiring survey under the BHP's survey priorities policy are designed to identify these types of sites, which have a high potential for containing information to address the research issues outlined above. In addition, even short-term, single-component sites without features can provide information on lithic material use and possibly on material culture, subsistence, and trade and exchange. As noted above, single component sites, though less abundant than multicomponent sites, are well-represented in the existing upland data and it is reasonable to expect that future surveys would identify this type of site.
A question raised by the survey priorities policy is whether multicomponent lithic scatters can address any of these research issues. Considering the data requirements discussed above, information from datable contexts is key. Defining chronologically distinct subassemblages of artifacts is problematical on multicomponent surface sites, especially those that have been plowed. In some cases, spatial patterning of lithic material types that predominated during a particular chronological period have been used to define temporally specific areas of occupation within a site (Bergman 1992; Miller 1995), but this approach only works in the few instances where lithic materials show strong temporal correlations and the conclusions must be considered tentative at best. In some cases, diagnostic artifacts are sufficiently patterned to suggest that different areas of a site were occupied at different periods. However, it cannot be assumed that all artifacts in a temporally specific subarea of a site were deposited during that time period; areas of occupation likely overlapped. Therefore, defining temporally specific artifact subassemblages on multicomponent sites cannot be done with a high level of confidence.
On multicomponent upland sites, features with wood charcoal generally represent the only datable contexts. Multicomponent sites with features have the potential to address some of the research issues outlined above. Preserved faunal and botanical remains from features can provide information on subsistence and on the seasonality of site occupation. Feature function provides information on site function. Patterns of features can provide information on social organization. Such sites should be, and generally are, considered National Register-eligible.
In addition to sites with features, the chronological distribution of site data indicates that any Paleoindian, Early Archaic, Middle Archaic, Early Woodland, and Middle Woodland sites would provide significant new information on settlement patterns in the three studies watersheds, and likely in the other low-priority watersheds as well. Any site dating to this period should be considered eligible for the National Register and intensively tested for the presence or absence of features. Unfortunately, so few sites from these periods are known that accurately predicting their environmental correlates, even using data from wider regions, is not possible.
Where are significant sites likely to be located?
One of the most interesting results of the project is the unique character of the site data for each of the three watersheds. After reading these reports, one has a very clear understanding of each watershed and the existing archaeological data. In addition to general research issues such as those listed above, watershed-specific research issues can be identified for each of these regions.
In general, the studies have demonstrated that the existing database in the three watersheds is not sufficient to understand upland settlement patterns. Few sites with chronological information are present in the data base, and most of those date to the Late Archaic and Late Woodland periods. The studies confirmed that most sites in the upland areas are multicomponent sites without features that will provide little additional data beyond their age and location. However, single component sites and multicomponent sites with features are more frequent than the overall PASS file data indicates and, because these site types have datable contexts, they can provide information on aspects of prehistoric behavior that cannot be interpreted using the existing data.
The studies have identified a number of apparent relationships between site location and environmental setting. However, because so few significant archaeological sites have been identified in the three watersheds, the environmental relationships are with all recorded sites, rather than with only significant sites.
•Proximity to water - Areas near first and second order streams are classified as upland in the survey priorities policy. The studies indicate that upland sites are more frequently located near these streams, as well as near springs. In the Conodoguinet watershed, 93% of upland sites are within 100 m of water and 83% were near water in the Brandywine watershed. The proportion of upland sites within 100 m of water was lower in the Conemaugh-Blacklick watershed (32%), possibly because so many of the sites were associated with lithic sources.
•Stream benches - A strong association between sites and stream benches was identified in the Brandywine watershed. More sites are located on stream benches than in other topographic settings, as are most sites where Phase II surveys were performed. The three sites considered significant or possibly significant were also located on stream benches. Th relationship between stream benches and site locations also appears to be present in the other two watersheds.
•Transportation routes - The studies also recognize that transportation routes, both in terms of known Native American trails and obvious topographic corridors, are correlated with site locations. Thus, areas near protohistoric trails, across low saddles, or along streams draining steep ridges tend to have a higher probability for prehistoric sites. This relationship was also identified in a settlement pattern study in the Bald Eagle Creek drainage, which is also one of the low-priority watersheds (Miller 1993).
Several other important insights have resulted from these studies. Perazio notes that upland sites in areas that have never been plowed or in areas that have been in long-term fallow have significantly greater integrity. Shallow features and activity areas are more likely to be preserved on such sites and fewer, if any, diagnostic artifacts have been removed by collectors. While such areas may not necessarily have a higher probability for containing sites, sites that are present would likely have greater research potential and therefore a greater probability of being significant.
Seigel et al. make an important point regarding site vs. non-site survey, specifically as it relates to the large number of low-artifact-density quartz scatters found in the Brandywine watershed and other areas of the Piedmont. These sites are widely distributed and show no strong environmental correlates. It may be that their function is such that environmental variables played no role in site location. In any case, the significance of these scatters is evaluated individually rather than as part of a general distribution of sites across a landscape. As a result, they are generally not considered significant. However, knowledge of the presence and distribution of these scatters is important in understanding prehistoric land use strategies in the Piedmont. Archaeological investigations in this province clearly need a landscape archaeology approach to interpretation of prehistoric settlement behavior as recommended by Siegel et al. While this factor does not in itself demonstrate that more sites of this kind are needed in the Brandywine watershed, it does highlight the fact that concepts of site significance need to be flexible enough to deal with the complexity of artifact distributions across the landscape.
Likewise, the results of the Meyersdale project summarized by Chiarulli indicate that surveys of large areas of the landscape provide us with an understanding of settlement patterns that is not inherent in the analysis of site data from small, disconnected survey areas. In large-scale studies, even sites that might not be considered individually significant contribute to our understanding of how prehistoric groups distributed their activities across a landscape. For this reason, we should not miss any opportunity for the survey of large, contiguous areas.
If the survey priorities policy is continued, it is recommended that it be revised to recognize the importance of surveys in unplowed or long-fallow areas, the importance of large-scale surveys, and the importance of a landscape, or non-site, approach for evaluating significance in some regions. In addition, given the distinct differences that have been identified in the environmen and the databases from the three studied watershed, additional studies of low-priority watersheds should be undertaken and directed towards the goal of developing watershed-specific management plans.
It is further recommended that the environmental settings summarized above be incorporated into the policy as high probability areas that should be surveyed in all of the 19 low-priority watersheds. However, it must be noted that by defining high probability areas for additional surveys based on the limited available site data and failing to at least sample low probability areas in future surveys, we run the risk of confirming our assumptions through biased research design. That is, the high probability areas that have been defined in the policy, or recommended for addition to the policy, represent hypotheses that must be tested. Given the severe limitations of the existing site data in the three watersheds, as detailed in this report, continuing the survey priorities policy without any testing of low probability areas could result in an increasingly biased data base.
The nature of the data on multicomponent surface sites means that research issues requiring tight chronological control on the artifact assemblage cannot be addressed. And although this site type is the most frequent type found in upland settings, single component sites, sites with features, and other sites that can address important research issues do exist. That more such sites are not represented in the existing data is the result of the fact that such small proportions of the three watersheds have been systematically surveyed. It is more or less inescapable that archaeological surveys in upland areas will have a relatively low return on investment, but this does not mean that the investment should not be made.
Given the findings in this project, the BHP may wish to consider replacing the survey priorities policy with an upland research protocol that would standardize testing procedures in upland settings throughout the Commonwealth and insure comparability among studies. The goal of the protocol would be to establish a database of systematic surveys that would lead to reliable predictive models based on sound data. When sufficient survey is completed in any watershed and a sufficient number of sites from all chronological periods have been identified, predictive models could be developed using the systematic survey data with GIS as a tool for analysis. The models could then be tested through surveys conducted as part of the Section 106 process. This process could conceivably lead to a point where we had sufficiently reliable predictive models that we could limit identification-level surveys in upland areas throughout the Commonwealth to high probability zones and be confident that significant sites would not be missed. In addition, this survey and testing process would provide useful information for interpreting settlement strategies in upland terrain.
In current practice, each researcher develops a predictive model for Phase I survey and in accordance with existing guidelines can test medium and low probability areas at a wider test interval than high probability areas. As a result, there is a greater likelihood of finding sites in high probability areas. The BHP protocol should establish a single shovel test interval for upland areas with low surface visibility and a single pedestrian survey interval for recently plowed fields.
The protocol should also include a standardized method of documenting the survey areas so that they can be entered directly into the GIS database. The reporting procedure could consist of a simple form and maps that would supplement the standard Section 106 reports. As a result, the GIS database would include accurate information on site locations and on settings that have been surveyed and found not to contain prehistoric sites.
The definition of when a sufficient level of survey has been completed to support reliable predictive modeling should be based on established statistical theory, which in this case is complicated by the fact that the selection of survey areas will be based on the location of compliance surveys rather than on statistical sampling. This determination should also consider the unique characteristics of each watershed and the specific research issues that should be addressed.
The studies described in this report have provided important insights into the shortcomings of the existing data in the three low-priority watersheds in terms of both accuracy and utility for archaeological research. Many of these limitations are the result of a lack of resources for field checking data and for undertaking the synthesis and analysis of regional data. Limitations also result from the fact that survey methods are generally developed to best meet the compliance goals of individual projects and, therefore, survey data is not always comparable among projects or across environmental settings. Development of an upland survey protocol would direct future surveys toward the goal of predictive models accurate enough to allow us to limit surveys to high probability zones with a high level of confidence. Clearly, our studies indicate that we have not as yet reached this level of knowledge.
|Table of Contents||Top of Page|
Bergman, C.A., P.C. LaPorta, J.F. Doershuk, H.A. Fassler, D.J. Rue and J. Schuldenrein
1992 The Padula Site (36Nm15) and Chert Resource Exploitation in the Upper Delaware River Valley. Paper presented at the 63rd Annual Meeting of the Society for Pennsylvania Archaeology, State College, Pennsylvania.
Bureau for Historic Preservation (BHP)
1996 The Development of Prehistoric Settlement Pattern Research Priorities in Pennsylvania. Pennsylvania Historical and Museum Commission, Harrisburg.
Duncan, Richard B. and Kristen A. Beckman
2000 The Application of GIS Predictive Site Location Models with Pennsylvania and West Virginia. In Practical Applications of GIS for Archaeologists: A Predictive Modeling Kit, edited by Konnie L. Westcott and R. Joe Brandon. Taylor and Francis, London.
Kvamme, Kenneth L.
1990 One-Sample Test in Regional Archaeological Analysis: New Possibilities through Computer Technology. American Antiquity 55(2):367-381.
Miller, Patricia E.
1993 Prehistoric Settlement Patterns in the Bald Eagle Creek Drainage of Central Pennsylvania. Ph.D. dissertation, Department of Anthropology, Pennsylvania State University, University Park, PA.
1995 Archaeological Data Recovery: Sites 36Sn220 and 36Sn221 and Additional Phase II Survey: Site 36Sn21, Safer Highways for Economic Development Project, S.R. 0011, Section 003 (T.R. 11 and 15), Snyder County, Pennsylvania. Report submitted to the Pennsylvania Department of Transportation, Engineering District 3-0, Montoursville.