Quantitative analysis of Munsell color data from archeological ceramics

Abstract

The Munsell color system is almost universally used for measuring colors of archeological artifacts. In addition to recording soil colors, many use the Munsell system to record the colors of ceramic attributes, such as pastes, slips, glazes, and paints. These data are often used in both modal and typological analyses of ceramic style. Most often, however, Munsell color data are used in a purely descriptive fashion, and the quantitative potential of the information is not fully exploited. We propose here a new protocol for manipulating, analyzing, and interpreting Munsell color data that permits the statistical investigation of sets of Munsell observations as well as hypothesis testing. A Munsell color reading is composed of three continuous interval scale variables, and these data can be transformed into x, y, z coordinates that define a location in color space (D'Andrade and Romney, 2003). Once transformed into spatial coordinates, Munsell data can be analyzed using spatial analysis techniques, such as k-means cluster analysis, as well as non-spatial statistical methods. In our case, we chose to use logistic regression to study the degree to which color could differentiate ceramic types and varieties. We argue that logistic regression is an appropriate approach for testing common types of hypotheses that archeologists may pose with ceramic color data. We illustrate the approach on sets of Munsell data representing ceramic slip colors from Mayapán, Yucatán, México. We were able to show a clear separation between the Mama Red and Polbox Buff types. Using the techniques we suggest here, archeologists can significantly expand their modal analyses of ceramics because color is a common and easily measured attribute in most ceramic assemblages. The same techniques can easily be extended to other kinds of artifacts, including lithics and textiles, as well as soils, sediments, plants, and other objects studied by field scientists.