SECTION 3.0
THE CONODOGUINET WATERSHED

3.1 Introduction

3.2 Research Design

The basic premise with which this study begins, one that is fundamental to modern anthropological archaeology, is that human behavior is patterned and that some of this behavior leaves discernable traces on the landscape. This is the basis of the predictive modeling that has become a basic part of much cultural resource management (CRM) archaeology (Miller 2000). As discussed below, a number of studies have achieved success in identifying statistical correlations between site distributions across the landscape and a number of simple environmental variables (e.g., distance to water, landform, soil type). These correlations are thought to reflect decisions made by prehistoric peoples regarding basic needs, such as food, raw materials, and shelter. Predictive modeling requires an adequate database on which hypotheses can be formulated and tested. If the available database does not support the identification of simple environmental correlations, then its adequacy for use in more advanced research is questionable. Such a finding would lead to the conclusion that basic data gathering must still be conducted within the given study area, and that relevant research questions must be framed so as to build the database both quantitatively and qualitatively.

Given the above, the overall goal of this study is to assess one of the BHP's excluded watersheds to determine the quality of the existing database and then to suggest some topics for future research appropriate to the current state of knowledge about this area.

The specific aims of the current contribution include:

• To assess the existing state of knowledge regarding the prehistory of the Conodoguinet watershed (Watershed 135), one of two in the central portion of the state that were selected by BHP has having as sufficient amount of existing data to merit exclusion of upland areas from further archaeological survey;
• To identify research questions relevant to this drainage that cannot be adequately addressed using the existing database; and,
• To make recommendations regarding modifications to the upland exclusion policy which may lead to the collection of additional data relevant to the identified research questions.

One other point must be made regarding the purpose of this study. Science proceeds by building upon itself. Answers to old questions lead to the formulation of new ones. Furthermore, there are many perspectives with which to view prehistory and generate valid, informative results. There is not a single, all-encompassing interpretation but, rather, multiple avenues to approach an understanding of prehistory. The present study will seek to identify some productive avenues for research. Many other important questions can be developed, now or in the future, and it is important to avoid, as much as possible, actions that preclude collection of the data needed to address such questions.

Assessment of the existing database was undertaken primarily using the resources available at the BHP. These consist of the computerized Pennsylvania Archaeological Site Survey (PASS) database, plus the paper records, including the site and project locations plotted on USGS quadrangles, PASS record forms, computerized PASS database, and CRM reports in the BHP library. It should be noted that not all relevant CRM reports could be located in the BHP library. These resources were supplemented by a variety of data sources regarding the environmental conditions within the drainage (e.g., topographic and bedrock geology maps, county soil surveys). Finally, publications dealing with regional prehistory and upland archaeology were consulted. The computerized database of CRM projects could not be made available by BHP.

3.3 Environmental Overview

Below is a brief overview of the environmental setting of the Conodoguinet watershed. As indicated above, an important aspect of the study of prehistoric human behavior is the attempt to understand the interrelationship between people and their environment. In order to properly assess the degree to which the existing database in this watershed can be used toward this goal, it is necessary to identify key aspects of the environmental side of the equation.

The Conodoguinet watershed encompasses 1,311.27 square kilometers (506.28 square miles) (Susquehanna River Basin Commission electronic data, n.d.). It extends from headwaters in Franklin County northeast to east through the northern portion of Cumberland County, ending at the Susquehanna River across from Harrisburg. There is a very small segment of the watershed in Perry County that has no recorded sites or surveys and will be excluded from the rest of the discussion. The Conodoguinet watershed is located primarily within the Great Valley Section of the Ridge and Valley Physiographic Province. However, the headwaters are located in the adjacent Appalachian Mountain Section of the Ridge and Valley. The Great Valley is a bedrock-defined landform which, in this vicinity, is demarcated by Blue Mountain to the north and South Mountain to the south (Figure 1). A number of gaps crosscut the divide between the Appalachian Mountain and Great Valley Sections.

Elevations within the watershed vary from approximately 122 meters (400 feet) at the confluence with the Susquehanna River, to 640 meters (2,100 feet) at the western end of the watershed. A low drainage divide on the Great Valley floor separates the Conodoguinet drainage from that of Yellow Breeches Creek to the south, also a tributary of the Susquehanna. To the west/southwest the Conodoguinet watershed is bordered by elements of the Potomac River drainage, and to the north by the Juniata drainage.

The headwaters of Conodoguinet Creek are located in a narrow, steep-sided valley between high ridges known variously as Blue Mountain and/or Kittatinny Mountain. The Conodoguinet enters the Great Valley through a gap just above the town of Roxbury (Figure 1). Within the Great Valley, the Conodoguinet watershed has a generally dendritic drainage pattern (Figure 2; Mueller 1993:1). However, there is a marked dichotomy between the tributary drainages to the north (and at the uppermost reaches of the drainage) and south of the creek. To the north, there are numerous, well-branched tributary systems. First-order streams tend to originate high on the valley wall, toward the crest of Blue Mountain, then progressively merge into higher-order streams as they descend ultimately to meet Conodoguinet Creek. Therefore, many of the northern tributary systems achieve third and even fourth-order status before reaching the Conodoguinet channel. By contrast, to the south, most notably below the confluence of Middle Spring Run, there are markedly fewer tributaries and these have fewer branches. The result is that to the south of the Conodoguinet there are broad, relatively level, inter-fluvial upland areas. This difference may, at least in part, be due to the topographic contrast between the two areas. Whereas the terrain to the north of the creek is largely a reflection of the relatively steep slope of Blue Mountain, to the south the Conodoguinet drains the northern portion of the Great Valley's relatively level floor. In addition, to the north of the creek and in the uppermost portion of the drainage, the bedrock is predominantly shale. To the south, below the Middle Spring Run confluence, limestone and dolomite are the predominant bedrock types (see below; Mueller 1993:1). Typically, more water travels underground in areas with the latter types of bedrock. This type of drainage and its attendant topography, which result from the dissolution of carbonate bedrock, are referred to as karstic.

Due to the highly branched pattern of the northern and upstream portions of the drainage, the Conodoguinet is already a third-order stream by the time it crosses from its headwater region into the Great Valley. Beginning at its confluence with Rowe Run, near Orrstown, Franklin County, it becomes a fourth-order stream. At its confluence with Middle Spring Run, near the Franklin / Cumberland County border it becomes a fifth-order stream. The Conodoguinet then retains this ranking all the way to its terminus at the Susquehanna River. Along the upper portion of its channel, the Conodoguinet forms relatively small meanders. However, in the lower portion of the drainage, beginning in the vicinity of Carlisle, the meanders tend to become larger and more closely packed as it cuts down toward the Susquehanna, which lies at a notably lower elevation.

The geological formations that make up the Great Valley portion of the Conodoguinet watershed are predominantly Cambrian and Ordovician in age (Berg 1980). Beginning at the crest of Blue Mountain and moving southward, these formations include Martinsburg (Om/Omgs), the Hamburg Sequence (Oh/Ohl/Ohg), Chambersburg (Oc), St. Paul Group (Osp), Rockdale Run (Orr), Pinesburg Station (Ops), Stonehenge Formation Stromatolitic (Os), Shadygrove (Csg), and Zullinger (Cz) (Figure 3). These are sedimentary formations, consisting of limestone, conglomerate, dolomite, chert, quartzite, and shale (Table 1). The three chert-bearing formations, St. Paul Group, Rockdale Run, and Shadygrove, would have been of particular relevance to prehistoric inhabitants of this region. All three are found to the south of the Conodoguinet on the gentle slopes leading up to the drainage divide with Yellow Breeches Creek. There do not appear to be any prehistorically significant bedrock sources of knappable stone in the Appalachian Mountain portion of the drainage (Table 1). Argillite and rhyolite are known in the watershed vicinity. Particularly important is the presence of known prehistoric (meta-)rhyolite quarries just beyond the southern margin of the Conodoguinet drainage (e.g., 36Ad153, 36Ad154, and 36Ad200). Knappable quartz and, to a lesser degree, quartzite occur in cobble form along Conodoguinet Creek and its terraces (Berge and Lewis 1993:23-26). One additional bedrock formation located in the Great Valley portion of the watershed is the Triassic diabase dike (Trd), which runs perpendicularly across the valley (roughly north-south) near its eastern end. This bedrock is correlated with the distinctive Athol-Neshaminy soil association (Group G, see Table 2, below).

Table 1: Bedrock Formations of the Conodoguinet Watershed

Landforms along the Conodoguinet Creek and its tributaries primarily result from fluvial erosion. Glacial erosion, periglacial mass wasting, and (as noted above) the dissolution of carbonates are also evident. "[Conodoguinet] creek has incised through beds of resistant shale and fine-grained sandstone leaving distinct escarpments between the residual uplands and the alluvial valley floor" (Sams and Beckman 1995:6; the distinction between alluvial and residual, vis-à-vis soils, is discussed at greater length below). As a result of this incision, the Holocene-age Conodoguinet floodplain is very narrow, between 2.0 and 10.0 meters (6.6 and 32.8 feet) wide for most of its course. It is "subject to frequent flooding, scouring, and redeposition of sediment" (Sams and Beckman 1995:6). The floodplain is poorly drained and/or deeply disturbed in most areas. The regular reworking of sediment, the limitations of its drainage, and the degree to which the creek has meandered have created one of the most geologically developed floodplains in the area (Berge and Lewis 1993:28).

Ecological variation within the Conodoguinet drainage may be broadly described using seven soil associations (Figure 4a [Cumberland County] and Figure 4b [Franklin County]; Long 1975; Zarichansky1986). These associations run generally west to east, paralleling the long axis of the drainage. Summaries of the relevant characteristics are presented in Table 2. To overcome the differences in nomenclature between Cumberland and Franklin counties, the associations have been cross-referenced and collective labels, termed 'Groups', have been assigned for convenience.

Generally speaking, the alluvial / floodplain contexts that the BHP survey priorities policy identifies as having continued research potential correspond with Group C soils. All other soil groups form on residuum, colluvium, or very thin alluvium, primarily along low-order streams. The upstream end of the Group C soil area is roughly at the point where the Conodoguinet becomes a fifth-order stream. The Group C area forms a relatively narrow band along the stream, which increases only slightly from its upstream to downstream end. It is notable that Group C soils do not extend to the Susquehanna confluence. As the latter is approached, the gradient increases and the stream channel more deeply incised, with little or no floodplain (Figure 1).

Excluding the Conodoguinet headwaters in the Appalachian Mountain Section, the upland portions of the drainage lie principally within two soil associations, Groups B and D, with a smaller portion in Group A. The Group B soils correspond roughly with the Martinsburg and Hamburg shale bedrock, primarily located to the north of the creek. Group A soils are at high elevation, arrayed along the crest of Blue Mountain, at the northern margin of the drainage. Group D soils are found on the limestone and dolomite bedrock to the south of the creek. Group E and F soils are found only in the very southern portions of the drainage, primarily at the southwestern corner. Finally, Group G soils are formed in the vicinity of a very narrow, north-south-running, diabase dike.

Table 2: Inter-county Concordance And Description of Soil Associations

Note: Definitions for Franklin County essentially similar to those for Cumberland County. Only differences are noted.

In general, the Group C and D soils are the most biologically productive, though many of the soil types in the other groupings are relatively well rated for their potential to support natural plant and animal communities which would have been important to prehistoric Native American populations(Zarichansky 1986:Table 9). Furthermore, as will be discussed more fully below, the highly branched pattern of tributary streams in the Group B area is likely to have created a large amount of ecotone, thus raising the biological diversity and productivity of this area (Odum 1971: 157). It should also be noted that karstic areas, soil Groups D and E, are likely to contain sinkholes, springs, reentrants, and closed depressions around which biological resources tend to concentrate (Lyman 1996:412-413; Waters 1992:215-219, 240-247).

Table of Contents

Top of Page

3.4 Review And Assessment of Existing Data

Archaeological Record of Watershed 135

When the BHP published its survey priorities in 1996, the digital PASS database listed 121 sites in Watershed 135. Our analyses of the Conodoguinet archaeological record began with this number. One site (36Da11) had to be discarded because it came from Dauphin County, across the Susquehanna in a different drainage. Three Historic-period sites were eliminated from our sample as they did not provide information relevant to the current research focus. When the BHP's paper files were consulted, we discovered that 36CU142 had been mapped in two separate--and very different--locations; this site had to be discarded because of the discrepancy. With these exclusions, our sample was reduced to 116 sites (Table 3). Given this sample, site density in the watershed can be calculated at approximately one site per 11.3 square kilometers (4.36 square miles). This density is only slightly higher than that derived by the BHP for the watershed (density = one site per 11.45 square kilometers), lower than the average listed for the state (density = one site per 8.77 square kilometers), and substantially lower than the average listed for the Great Valley (density = one site per 3.90 square kilometers) (BHP 1996:16-17, 30).

The 116 sites of the Conodoguinet sample (like those of other watersheds in the state) derive from two sources: (1) unsystematic collections by avocational archaeologists throughout the drainage; and (2) systematic collections by archaeological professionals at very circumscribed locations within the drainage (Table 4). In 1996, only about 88.3 square kilometers (34.1 square miles; 6.08 percent) of Watershed 135 had been subject to any sort of professional archaeological examination. For the most part, this total represents the efforts of people conducting informant interviews and background research to create predictive models of archaeological potential across large portions of the drainage.

Table 3: List of Sites in Watershed 135 Showing Setting Criteria

^(b) Rockshelter / reentrant (special setting)

Note: 4 sites could not be classified in either category.

Table 4: Method of Site Discovery by Setting

Problems with the Database

Biases in the PASS data for the Conodoguinet drainage derive from a number of sources. Some, like those noted above, are simple clerical errors. Our need for standardization and dependence upon reference codes in databases make such errors common. For example, sites are easily listed in the wrong drainage through missed key strokes. (Our analysis of the Bald Eagle drainage [Watershed 190] data turned up a number of sites listed in the wrong basin: sites clearly located in the Susquehanna basin were tabulated in the Delaware and Allegheny River valleys.) These factors also come into play when a number of people are entering data into a single digital file. Without strict control over the reference codes there is bound to be some variance in their entry. This is particularly true of databases (e.g., the PASS files) that are being generated over a significant period of time. The mapping of the same site in two separate places is less easy to explain, but forgivable. Clerical errors like these can be remedied through a thorough review of the data. Such a review will undoubtedly have to take place before the BHP can complete its plans for the generation of a state-wide GIS database (van Rossum and Pollack 2000). It should also be noted that the data available in the computerized PASS database has not been consistently updated since 1996. Newer data reported in this study were obtained directly from the paper files (i.e., PASS forms, USGS quads, and CRM reports) in the BHP library, but were not included in the statistical analysis.

Other biases are less easily remedied. In a number of cases, data for particular sites are absent from the PASS database. These data might be single variables or entire sets of information. When the original PASS forms are consulted, it is often the case that the data were never listed. Thus, a few sites in Watershed 135 exist in trinomial number and artifact alone. This state of affairs has hindered earlier attempts at synthesizing the Conodoguinet data (Berge and Lewis 1993). It can only be assumed that the rest of the database is equally flawed. Recognizing that ours is a destructive science, we have no way of knowing what data these sites contained prior to their excavation / collection. While it is easy to point fingers at the (often) less-trained avocational archaeologist, the responsibility for this information belongs to all of us.

In trying to recover unlisted data from site reports, we discovered another bias. The field "VEGETATE", which is supposed to list codes that describe the site's surface condition / cover (e.g., active agricultural, pasture, thicket and shrub, forest, etc.), was blank for almost every entry in the Watershed 135 database. It took a review of the PASS forms and the site reports to determine that most of the sites had been plow disturbed (a characteristic noted by the BHP for most upland sites [1996:6]). Further analysis, however, revealed that the inordinately high number of plow-disturbed sites in upland--as well as riverine--contexts was not necessarily reflective of diagenetic factors. Rather, the number resulted from recovery techniques. In a number of cases where project area contained both agricultural and wooded areas, the BHP only required surface collection of the freshly plowed field (see Kinsey 1992a, b; and recently Geidel 1998). These surface collections were occasionally supplemented by subsurface testing, but rarely in the wooded areas. This recovery bias has led some researchers to conclude that "large and significant prehistoric sites are located along the floodplain of Conodoguinet Creek" and that "the interior and upland settings. . . are low probability areas for site locations" (Kinsey 1992a:11). To these researchers "most upland sites are short-term hunting and tool maintenance sites" (Kinsey 1992a:11) with limited research potential; a view which the BHP (1996) maintains. However, this interpretation is circular, since it is based on data derived overwhelmingly from plow-disturbed sites with wooded contexts being largely excluded. The question remains, "What really lies below the ground in the wooded areas?"

Additional problems were encountered with the PASS database for Watershed 135. These problems related to specific queries that we made of the database as we attempted to analyze the data. These problems will be described in the relevant sections below.

3.4 Review And Assessment of Existing Data -- continues

Table of Contents

Top of Page