A general outline of, and justification for, this study, as well as a discussion of the major related concepts, are presented in this chapter.

Carl Lloyd supervised an archaeological survey of the area adjacent to the Ackmen pueblo in southwestern Colorado in 1937 in order to supplement the data recovered from the excavation of this famous site. Sixty-six quarter-sections (each equivalent to a ¼ sq mi area) measuring ½ mi on each side, were intensively surveyed by a large crew. Members of the crew were stationed 100 ft apart along the ½-mi width of the quarter-section, and the entire crew walked across the ½-mi length of the quarter-section.

From each section consisting of 4 quarter-sections, 2 diagonal quarter-sections were selected for investigation. Lloyd assumed that 66 surveyed quarter-sections would reflect adequately, at least, on a 33-sq mi area near the rim. This assumption was based on the rather large (50%) sampling fraction that he employed. He was not certain that his survey methods would produce a contribution to the archaeology of the area and stated the uncertainty as follows: “Given a discontinous intensive archaeological survey of a region, what contributions can be made to the archaeology of that region by means of an analysis of the data thus obtained?” (Lloyd 1938:282). He did not explicitly answer that question during the course of his report. However, I believe that his answer was implicitly positive, that is, the kind of survey he performed did produce a meaningful contribution to the archaeology of the Ackmen-Lowry region.

The Paria Plateau survey was conducted on, and adjacent to, the Paria Plateau in Coconino County in north-central Arizona by personnel of the Museum of Northern Arizona during the summers of 1967 and 1968 (Fig. 1). All of the surveyed area is located in the Arizona Strip District of northern Arizona, although unsurveyed portions of the Paria Plateau extend into southern Utah. The Plateau is situated approximately midway between Page, Arizona on the east, and Kanab, Utah on the west. The familiar Vermillion Cliffs seen from U.S. 89 enroute to the North Rim of the Grand Canyon form the southern escarpment of the Plateau. The western boundary of the surveyed area is defined by Coyote and House Rock valleys which also separate the Kaibab and the Paria plateaus. The Paria River constitutes the northeastern boundary of the Paria Plateau making it a naturally bounded region within the Colorado Plateau subarea.

The remainder of this study is designed to determine how a hypothetical region that is archaeologically unknown must be surveyed in order to derive valid conclusions concerning the entire region. This sampling problem can be fruitfully approached by performing an experiment with the data recovered during the Paria Plateau survey. An experiment is necessary because the stated purpose of this study is to refine sampling methods in an empirical manner. Therefore, a maximum of data that can be handled properly under the correct theoretical conditions can be used to answer questions concerning the sampling methodology of archaeological survey. Second, I consider this study experimental in an innovative sense; to my knowledge, the kinds of analyses that will be described in this chapter have only been partially undertaken by other archaeologists (Chenhall 1972; Plog 1972; Judge, Ebert, and Hitchcock 1973).

The data generated by the methods described in Chapter III are summarized in this chapter in terms of the problem of this paper. As stated by Lloyd, the problem concerns the validity of conclusions based on a discontinuous intensive survey. In this study, 326 discontinuous intensive surveys are simulated on paper, and the results are interpreted in terms of the 3 aspects of the sampling process–sampling scheme, fraction, and unit. Which alternative of each aspect of the sampling process is more likely to produce valid conclusions? I hope that an empirically based answer to this question will assist archaeologists who want to survey complete regions, but are forced to sample because of temporal and financial limitations.

Chi square and economy are useful for evaluating each simulated sample and for comparing each sample to the population. As outlined in the previous chapter, the experiment generated 326 samples. For each sample, a chi-square value for each of 5 variables and 1 number representing economy are calculated. Thus, the data used to answer the question of this study consist of 1,680 chi-square values and 326 values of economy.

In this experiment, the frequencies of occurrence of prehistoric cultural traits have been used to evaluate various alternatives of the process of sampling a hypo the tic ally unknown region. The sampling process is manifestly useful in archaeological survey.

The experimental results are only as valid as the methods, data, and theory on which the results are based. The sampling theory has been presented in Chapter III. Therefore, a synopsis of the research methods of the Paria Plateau survey is presented in this section so that the reader can continue to evaluate the data and the results of the experiment.

As frequently happens in contract archaeology, the topic of this experiment was formulated after the completion of the fieldwork. Therefore, the experiment depends on data that were collected to solve non-sampling problems. If the problem of this study had been conceived prior to fieldwork, it would have been possible to collect interval level data (such as the area of the site, of surface refuse, and of dwelling units) and to perform controlled, probabilistic surface collections. This would have added 2 interval variables—artifact and debris densities—and allowed for the employment of more powerful statistical tests.