The Archeological Survey: Methods and Uses

Cover

Contents

Foreword

Acknowledegments

Introduction and Definitions

A Brief History of Archeological Survey

The Variety of Archeological Survey

Basic Archeological Site Survey Methods

Special Types of Survey

Recording and Reporting

Predictive Survey for Comprehensive Planning

Conclusion

Bibliography

Forms Used in Recording Archeological Survey Data

Archeological Predictive Studies

Example of an Archeological Review Procedure Using Predictive Data

Automated Management of Data and Research Results on Archeological Surveys

State Archeological Co-ops: Their Evolution, Dangers, and Value

The Archeological Survey: Methods and Uses

Much of the recent archeological literature concerning archeological survey has dealt with the conduct of predictive survey--i.e., those that result in the prediction of archeological site distributions within a large area based on a less than complete survey of the area. Most of this literature reports studies conducted for research purposes in the arid-semiarid west (cf. Gumerman 1971; Thomas 1969; Mueller 1974, 1975). Lovis (1976) has reported a predictive survey for research purposes in the northeastern woodlands. Agencies with broad land-management responsibilities have learned the advantages of using predictive surveys in their general planning activities; as a result both the U.S. Forest Service (Smith 1977) and the Bureau of Land Management (Weide 1974) have begun to develop predictive survey programs. Both the U.S. Army Corps of Engineers (Schiffer & House 1975) and Interagency Archeological Services (King & Hickman 1973) have sponsored predictive surveys in order to determine the indirect or secondary effects of public works programs. State Historic Preservation Officers have begun to undertake predictive surveys in connection with their statewide comprehensive survey and planning programs (cf. Illinois 1977). Interagency Archeological Services has sponsored pilot studies that attempt to predict the distribution of prehistoric sites in two terrestial areas (Dincauze & Meyer 1976; Benchley 1976) and both historic and prehistoric resources on the Gulf of Mexico outer continental shelf (Gagliano 1977), for multi-agency management purposes. Appendix B lists a number of exemplary predictive studies.

The purpose of this brief chapter is to discuss very generally how predictive sample surveys are done and how they can interlock with programs of comprehensive planning--particularly those programs of statewide survey and planning undertaken by SHPOs.

What is Predictive Survey?

For purposes of planning, a predictive survey can be viewed as an attempt to build a data base for sensitive, responsible historic preservation planning without conducting a 100% non-exclusive survey. In a predictive survey one physically inspects only a fraction of the actual area of concern, and from this inspection--in the context of good background research--extrapolates to the entire area. Based on a predictive survey of high reliability (i.e., one that has been subjected to a number of carefully planned and executed tests) it should be possible to conduct controlled-exclusive surveys instead of non-exclusive surveys in advance of construction or land-use projects, thereby saving a considerable amount of time and expense.

Although predictive surveys can be useful in the preliminary analysis of relatively small areas (cf. Miller 1975), they are usually most cost-effective when applied to large regions such as States, river valleys or mountain ranges.

Research Design in Predictive Survey

A good research design is vital to the success of a predictive survey. The design should specify the general types of properties that will be sought and the criteria by which they will be evaluated. This specification should not only recognize current archeological (and other) research needs that ascribe value to various types of sites (cf. King 1971; Schiffer & House 1977); it should also seek to categorize sites into descriptive types so that an attempt can be made to preserve a representative sample for future research (cf. Glassow 1977a, 1977b; Glassow & Spanne 1976; Hickman 1977; King, Hickman & Berg 1977). In addition, the design should set forth a strategy or strategies for sampling the study area. Such a strategy should provide a data base for predictions that are rigorously testable (cf. Mueller 1975; Smith 1977). Only when a set of predictions has been tested and shown to represent accurately the actual distribution of archeological sites, can it be used confidently in planning. Each new survey should be designed in such a way as to constitute a test of the predictions until a high level of confidence has been reached.

Development of a research design for statewide comprehensive survey, or for elements of such a survey, should involve scholars having research competence in the study area, in related areas, or in general anthropological or other research pertinent to the study area. The research design is an appropriate part of the State plan for historic preservation, and in fact might sometimes constitute the State plan with respect to archeology (see Michigan 1977, Georgia 1977).

Background Research in Predictive Survey

Background historical, archeological, and environmental research overlaps considerably with research design formulation and may be necessary before the research design can be fruitfully prepared. Background research provides a basis for the first stage of the predictive study; based on background data one attempts to predict what kinds of sites will occur in the study area and where they will be found.

Studies of historic land-use patterns should enable one to at least predict general relationships between human activities and aspects of the natural environment (see Hickman 1977 and Bettinger 1977 for examples). For example, background research in the general vicinity of Griffin Valley might result in the development of a table of predictions similar to that shown in Figure VII-1. The same kind of predictive tables could be developed for prehistoric site-environment relations, as shown in Figure VII-2. Stratification of the sampling universe (discussed below) could then take these predicted relationships into account and fieldwork would test their accuracy.

FIGURE VII-1.

PREDICTED TYPES AND DISTRIBUTIONS OF
HISTORIC SITES: GRIFFIN VALLEY AREA

FIGURE VII-2.

PREDICTED TYPES AND DISTRIBUTIONS OF
PREHISTORIC SITES: GRIFFIN VALLEY AREA

It is important to emphasize that predictions from background research should be based on accurate settlement pattern data and well-grounded hypotheses about human-environmental relationships insofar as possible. An alternative approach is to take data on the distribution of known archeological sites relative to features of the environment and then project a similar distribution of sites relative to all similar environmental elements in the study area. Benchley (1976) and Dincauze & Meyer (1976) have done this for the Greater St. Louis and Eastern New England areas, respectively. There are great dangers in this approach, however. Because we typically have no idea how information on known archeological sites in any area was gathered (remember Beakey) it is seldom possible to state how representative the distribution of such sites may be of the actual distribution of all sites. It is safe to assume that the answer to this question, in most cases, will be "not very." In many instances, known archeological sites are found to be consistently associated with modern roads and highways; this does not reflect the habits of historic and prehistoric people nearly so much as it does the habits of archeologists. Were we to predict the distribution of sites in Griffin Valley from Beakey/Loumington data our result would be a map like the one shown in Figure VII-3--not inaccurate in its plotting of "high probability areas" as far as it goes, but leaving many areas of actual high sensitivity designated as "low probability areas."

Regardless of the sources used, predictions based on background research without field verification should never be used as the sole basis for planning. In other words, one should never assume that because unverified predictions from background data indicate that an area has low archeological sensitivity, a construction project planned for the area will not need to be surveyed. There are necessarily some pragmatic exceptions to the rule. On the outer continental shelf, field verification of predictions is extremely difficult, and more reliance on background data than would otherwise be acceptable may be necessary (cf. Gagliano 1977). Similarly in urban situations it may be necessary to rely more heavily than usual on background data in order to avoid very costly and difficult subsurface exploration in "low probability" areas (see chapter V above). In all cases, however, as much field verification as possible should be completed before predictions are used as planning tools.

Fieldwork in Predictive Surveys

On the ground the same methods are used for predictive survey as for any other kind of survey but only portions of the whole study area are actually inspected. It is vital that the portions selected be representative of the whole. Much of the literature on predictive survey deals with the problems of choosing a sample to inspect that truly represents the important aspects of diversity within the environment of a study area.

Choices of sampling scheme: The system used to select the portions of the study area to be inspected is called the sampling scheme, and several types of schemes have been or are being used. The simplest scheme is that referred to by Mueller (1974:39) as the "grab sample," in which one simply "grabs" whatever data are available and makes predictions from them. We have discussed this approach above; because it is not really sampling at all, it is not recommended except under particular circumstances. In one instance, this author felt justified in using a "grab sample" together with a considerable amount of background data to generate predictions. In and around the city of San Jose, California, as a result of a State environmental law and local implementing regulations, most private housing projects and other programs of public and private construction were (and are) subjected to prior archeological survey. Following guidelines set by the State archeological society (Society for California Archeology, 1973) most survey reports were filed with a local "archeological clearinghouse." In general, the surveys had been conducted in accordance with consistent fieldwork standards. As a result, after about two years during which the environmental law had been in force, the clearinghouse had data on a large number of small surveys scattered over the entire area of the city and its environs, including most natural environmental zones. These data were "grabbed" and used as a preliminary test of predictions generated from background research (King and Berg 1974). Such a test can only be regarded as preliminary, however, because of the uncertain representativeness of the sample and the possibility that field techniques used were not uniform among the various surveys.

"Simple random sampling" is a method of eliminating the biases from sample selection. The study area is first divided into equal-sized units (e.g., quarter-sections) which are assigned numbers. Units are then selected at random using a table of random numbers or some other objectifying device. Simple random sampling is useful in a homogeneous study area in which one lacks a basis for recognizing different subareas on environmental or other grounds. If the area contains different environmental zones that may have influenced settlement patterns, or if background research has suggested that particular subareas may contain particular types or densities of sites, simple random sampling essentially wastes this knowledge. Even if it is known that a given subarea is likely to contain a particular type of site, one cannot go and look unless the table of random numbers so directs.

"Systematic sampling" has similar weaknesses. In a systematic sample units are chosen at regular intervals on a grid so that one obtains a sort of checkerboard effect. Although this approach distributes sample units well over the study area, it does not guarantee that all subareas that may be of interest are sampled.

"Stratified random sampling" has become the most common general sampling scheme used in archeology. In stratified sampling one first recognizes and delineates those subareas that are thought likely to contain different kinds or densities of sites or those attributes of the environment that are thought likely to have influenced settlement patterns. This, of course, is done on the basis of background research. Next, each subarea, or "stratum," is divided into units, a sample of units is selected at random from each stratum, and the size of the sample is weighted to assure equivalent representation from all strata.

Figure VII-4 contrasts the kinds of samples that might be selected from Griffin Valley using simple random, systematic, and stratified random sampling schemes, assuming that environmental zones were used to define sampling strata in accordance with normal practice (cf. Thomas 1969, 1973; Mueller 1975).

Figure VII-4a. Simple Random.

Figure VII-4b. Systematic.

Figure VII-4c. Stratified Random.

A variant on stratified random sampling, which takes logistics into account, is referred to as "cluster sampling" (cf. Matson and Lipe 1975). Once the study area has been stratified, clusters of potential sample units are defined with each cluster designed to include representatives of all possible strata. Sample units are then selected within each stratum within each cluster rather than being scattered throughout the study area. This approach is obviously more cost-effective than straightforward stratified random sampling in most cases.

Choices of sampling unit: Sampling units are the parcels of land chosen for inspection. Both unit size and unit shape are important because the units must be comparable and because their size and shape affect the number of units that can be selected and the efficiency with which they can be inspected. Archeologists have traditionally used "quadrats" as their sampling units. Quadrats are squares of some convenient size (often a quarter-section), selected from a grid superimposed on a map or airphoto of the study area. The sample units shown in Figure VII-4 are quadrats. Many archeologists are now switching to "transects," which appear to be more effective and less costly to inspect. A transect is a long, thin rectangle, perhaps a mile long and 100 meters wide. They are laid out in such a way as to cross-cut sampling strata, thus allowing the surveyor to observe greater variability than is possible with a quadrat; they are also easier and faster to cover on foot and can be inspected with fewer people. Figure VII-5 shows a stratified random transect survey selection at Griffin Valley.

Figure VII-5. Stratified Random Transects.

Choices of sampling fraction: The sampling fraction is that portion of the total number of sampling units in the study area that is chosen for inspection. Sampling fractions used in predictive surveys have ranged from under one percent of the whole study area to over 50 percent (Mueller 1974:30). No reliable estimate can be made about the fraction needed to produce satisfactory predictions. Archeologists in the southwestern United States hold opinions that range from 15 percent (Donaldson 1975:15) to about 40 percent (Mueller 1974:66), and the one available study from the northeast estimates that around 20 percent is adequate (Lovis 1976). Obviously the exact size of the sample necessary depends on the nature of the data needed (cf. Read 1975), which in turn depends on the nature of the archeological sites in the area, the nature of the area itself, and (for our purposes) the management needs that prompt the study. As a rule of thumb, we suggest that for a large area like a State, a rough idea of site densities can probably be obtained from a sample of about one percent, combined with thorough background research. To obtain finer-scale predictions, define site-types, or make predictions about smaller areas, much larger samples are needed. It is important to remember that the purpose of sampling is not the discovery of sites but the establishment of expectations about where sites will be and what they will be like. Sampling is not a substitute for complete survey but one step in the survey process.

The Results of Predictive Survey

Predictive survey results are usually presented as maps portraying differential site densities or, in some cases, site type-densities. Figure VII-6 is a predictive map of a large area in the State of Indeterminate, including Griffin Valley, in which predictive survey has advanced to a level at which both site densities and site types, including both historic and prehistoric sites, can be predicted. Predictions are made by noting the association of particular site-types with particular environmental features within the sample units actually inspected, then projecting a similar distribution of sites to all equivalent environmental features within areas not yet inspected. An association gains reliability if it was first predicted on the basis of background research and later verified by field survey. An association also gains credence if it is explicable once discovered (e.g., the camps of unrecorded gold miners are found near placer deposits), and if it is consistently re-verified by further testing.

Figure VII-6. Predictive Map: Probable Distribution of Archeological Sites: Cooper-Cole Water Pollution Control District.

Because many associations of sites with environmental features are relatively uncomplicated, many predictive maps have simply been prepared by hand (cf. King and Hickman 1973). For plotting more complex relationships, and applying statistical tests to associations, analysis and preparation of maps with the aid of a computer is advisable. SYMAP (synagraphic mapping) is a program designed to prepare such predictive maps (Dougenik and Sheehan 1975).

Hierarchical Predictive Surveys

For comprehensive planning, predictive survey may best be considered an ongoing process in which increasingly fine-tuned predictions can be made as more and better information becomes available. If the archeologist continues to survey a new selection of sample units every year, he will eventually obtain a 100% sample. This is a rational goal for statewide comprehensive surveys and for Federal agency surveys conducted under section 2(a) of Executive Order 11593. The advantage of predictive survey is that some useful data for purposes of planning in the entire study area become available almost immediately,--for many parts of the country at least--and it is probable that all the information needed to carry out responsible preservation will be available before physical inspection has covered even 50 percent of the land.

The ongoing predictive survey process is best conceived of as hierarchical, with refined predictions being built each year (or some other appropriate interval) through testing of older, less certain ones. The following basic phases can be projected:

Phase 1: Background research serves as the basis for developing preliminary predictive tables and maps. Field survey strata are established and a sampling scheme is developed.

Phase 2: Initial sample fieldwork is undertaken. Depending on such factors as the size of the study area and the level of funding the sample fraction might be less than 1 percent or it might be as much as 10 percent; it will provide a rough check on the predictions developed from background research and result in a fact-based but still general prediction.

Phase 3: The sampling scheme is refined based on the results of Phase 2. The sample fraction can be increased with further fieldwork and new background research may be undertaken to seek information on phenomena identified in the field. Results should include more refined predictive maps and data.

Subsequent phases will further refine the survey results on the basis of increased sampling fractions and testing of the results of previous phases. When additions to the predictive maps become entirely repetitive i.e., when predictions are consistently verified by subsequent fieldwork, the predictive survey can be terminated. From this point on, barring the discovery of new types of sites or other newly important forms of data not previously attended to, only controlled-exclusive surveys would be necessary in the study area in advance of projects affecting unsurveved lands.

Crisis-free and Crisis-oriented Surveys

When developing a predictive survey program for a large and complex area like a State, the responsible officials will need to choose between crisis-free and crisis-oriented strategies. A crisis-free strategy is one in which present or possible future land-management problems in particular portions of the study area are ignored, and all portions are given equal representation in survey design. Conversely, a crisis-oriented survey is one in which attention is focused first on portions of the study area where preservation crisis conditions exist or are expected. Crisis-oriented strategies are perfectly reasonable, and are probably the most reponsible strategies for SHPOs to adopt in most cases. In deciding between a crisis-free and a crisis-oriented strategy, however, a SHPO should consider the whole range of crisis-types that may confront preservation in the area. For example, the most obvious kind of crisis may be the destruction of prehistoric sites in one part of the State through reservoir construction. But this may be obvious only because such projects appear daily on the SHPO's desk through the A-95 Clearinghouse; on the other side of the State important sites may be lost to landlevelling in connection with agricultural expansion, unknown to the SHPO because A-95 procedures do not apply. Although it may be quite possible to handle the impacts of the reservoir construction through existing procedures for compliance with the National Environmental Policy Act, Executive Order 11593, and other authorities, the agricultural damage constitutes a more serious crisis because of its unregulated nature.

Applying Predictive Survey Data

The data produced by predictive surveys can be integrated into the activities of State and Federal agencies for pure preservation planning and to facilitate compliance with the National Environmental Policy Act, Executive Order 11593, and other preservation-related authorities.

Realizing preservation opportunities: Predictive maps can be used in planning for the preservation of open space and for the acquisition of parkland. Areas of predicted high archeological site density, or areas where particularly significant types of sites can be predicted to occur, can be identified as high priority areas for public acquisition, open-space zoning, or other forms of protection. Such data can also be used by owners and developers of private land to guide development in such a way as to preserve archeological values.

Local regulation of land-use: Many States now have environmental statutes that require consideration of historic and archeological values in local general planning and in granting local permits for private land-use. Local and State plans (e.g., Coastal Zone Management Plans) assisted by the Federal Government must take these values into account even if State law does not require it. Predictive data can be of great aid to local planners and decision-makers by helping them decide when private applicants for permits should be required to undertake archeological surveys.

A-95 and environmental review: SHPOs who participate in environmental review under the procedures set forth in the Office of Management and Budget's Circular A-95, the procedures of the Advisory Council on Historic Preservation, or the procedures of other agencies, will find predictive data very useful in making responsible comments on project proposals. Appendix C is an excerpt from the review procedures that have been employed by the North Carolina SHPO that indicate how predictive data can be used in environmental review. During the early phases of a hierarchical predictive study, when predictions are still tentative, it will be important to be relatively inclusive in calling for survey data before commenting on a project because the predictions cannot be relied upon. Surveys that are done as a result of the environmental review will contribute to testing the predictions, however, and help allow the SHPO to narrow the range of projects for which he will request surveys in the future.

NEXT >>>

Last Modified: Mon, Jul 25 2005 10:00:00 pm PDT
http://www.cr.nps.gov/history/online_books/king/chap7.htm