Archaeologists have been using informal, very humanistic sampling (that is based on neither statistics nor probability theory) since the beginning of the profession. A vast body of archaeological knowledge has been built on this approach in the formative decades of the profession. In general, the emphasis in the body of knowledge has been on the spectacular and unusual finds, which constituted romantic conceptions of archaeology, and the opportunity to find what no else has seen or touched in hundreds or thousands of years. In addition to the phrase, intuitive sampling, other terms have traditionally been used to refer to related variations of haphazard, nonprobabilistic sampling. These terms include purposeful, typical, judgmental, and grab samples. Purposeful samples might be those intentionally selected by the archaeologist to hit something cultural, for example, the top of hill where a site/ scatter might be expected. Typical or judgmental samples are those areas targeted to recover typical remains or to get at high-density artifacts scatters. At the other end of the intentionality spectrum, grab or haphazard samples are more or less thoughtless, data-free samples selected spontaneously and that continue to be used in spite of late-twentieth-century archaeological scientism. Investigations conducted by these means are part of what has been called descriptive surveys. Samples gathered by any of these informal means have very restricted analytical value, and the corresponding statistics can only be used to describe the data in the sampling units from which data were collected. Statistics applied to such samples cannot be used to generalize to any larger group, not to the larger population of which the sample is only a small part (see Statistics in Archaeology).
However, in the second half of the twentieth century, biases and distortions in the interpretation of the archaeological record were realized. In that time period, new theoretical approaches (i. e., paradigms) to correct those distortions arose as scientism crept slowly, and with much internal professional resistance, into the practice of archaeology. Environmentalism, public archaeology, postmodernism, computer applications, systems theory, and model building are parts of the paradigm shifts. The introduction, thoughtful/ reflective use, and in-progress evaluation of many kinds of sampling has been part of the new methodological rigor and postmodern paradigm shifts. In the 1960s, the formative years of the shift, archaeologists used the sampling literature of sociologists and cultural geographers as guidance for developing parallel applications of sampling to archaeological investigations. This dependence is a continuation of the trend in American archaeology’s parent field, anthropology, that has a centuries-old tradition of borrowing concepts from related social sciences.
For many years, the various kinds of probabilistic sampling have been traditionally identified by their scheme, for example, simple random (SRS), systematic, stratified, and cluster (to be described later herein). Recently, these traditional schemes have been grouped and labeled as conventional designs or design-based approaches. Adaptive and nonstandard designs (bootstrap and jackknife subsampling) were added to the sampling repertoire of statistical theory and social science beginning in the late 1970s. These new statistical additions are finding their way into archaeology. They may be considered more user-friendly to the nonquantitative archaeologist who feels peer pressure to employ probabilistic sampling. Opposed to those conventional, design-based approaches are model-based approaches in which a constructed model is used to provide assumed, partially empirical information about the population being researched.
The distinction between probabilistic and nonprobabilistic sampling has weakened since about 1980 in the intellectual postmodern era, as has the traditional emphasis on sampling without replacement. Jackknife and bootstrap subsampling are two techniques in this intellectual gray zone between probabilistic and nonprobabilistic sampling. In a word, jackknife and bootstrap subsampling are means of secondarily studying the behavior of selected samples, are statistically similar to sampling with replacement, and are needed for estimating population parameters more accurately. Bradley Efron contributed to the description of these techniques. They are suited for complex sampling schemes, for example, the multistage cluster, complex stratified schemes, and combinations of both cluster and stratified. Jackknife and bootstrap subsampling seem to be statistical solutions to the archaeological research paradox. This paradox says that empirical information about the ‘substantive elements’ of research interest (sampling elements, e. g., sites, features, artifacts, household room blocks) can only be found by investigating artificially contrived, methodological units of observation, that is, sampling units (cf. ‘Glossary’). This artificiality, this paradox creates a vast disconnect between archaeological research and statistical theory and has been described by Mueller as “Archeological Research as Cluster Sampling’’ and by Orton. More research is needed in this area.
World History
