Interpretations

It is important to see interpretation of data as different to data analysis. Data analysis merely arranges the data in a more understandable form (or sometimes tests the statistical ‘significance’ of the patterns observed, i.e. their likelihood to be due to chance). Interpretation is still required to assign meaning to the patterns identified. Data do not speak for themselves nor is data analysis a ‘black box’ which takes data in at one end and produces an interpretation at the other. When interpreting patterns in archaeobotanical data, it is usually necessary to incorporate other information about the archaeobotanical samples (date, context etc.) and/or information about the botanical taxa (ecology, physical characteristics etc.). Interpretation also relies on experience of the present as it is only by reference to the known (present) relationships between actions and data that meaning can be assigned to archaeological data, where the actions which generated them are unknown.

Some data analysis methods, such as principle components analysis (PCA) or correspondence analysis (CA) are ‘pure’ pattern-searching techniques in the sense that they do not use, at the analytical stage, any information concerning the archaeobotanical sample, or the taxa concerned, other than the numerical composition of the samples; samples are arranged in relation to one another purely on the basis of botanical composition. To interpret these patterns, additional information can be used to ‘interrogate’ the PCA or CA plot. For example, the sample data points in a correspondence analysis plot can be coded according to archaeological period/phase, or context, to determine whether the compositional patterns are related to chronological changes or different activity areas, for example changes in the crops cultivated through time or differences in the use of different structures.

Similarly, the taxa in the samples may be classified according to their physical characteristics (seed size, shape etc.) or ecological preferences/characteristics (habitat preferences, life cycle, ecological traits etc.) to determine the underlying reasons why the species composition of some samples is different to others. For example, differences in weed seed characteristics may relate to crop processing activities associated with different archaeological contexts, while differences in weed ecology between different phases may relate to changes in cultivation methods. Similarly, ecological groups of wild species have been used to indicate the very existence of cultivation, as at Tell Mureybit, Syria, where 'pre-domestication cultivation may be indicated prior to the appearance of domesticated crops (Colledge 1998).

These types of post-analysis investigations therefore assist the interpretation of plots resulting from data analysis. Such interpretations are also based on a knowledge of the expectations (in terms of the types of taxa represented) for different crop processing stages, cultivation practices etc. These expectations can be generated only in the present, where the relationship between botanical composition and human actions can be observed. These observations can then be extrapolated, with due caution, to archaeobotanical compositional patterns in order to identify the past actions that gave rise to the archaeobotanical remains. Present-day ethnographic and ecological studies therefore play a significant role in generating these expectations.

An alternative way of using ethnographically or ecologically derived expectations in the interpretation of archaeobotanical data, is to incorporate an element of interpretation into the original data analysis. One way of achieving this is through discriminant analysis, using ethnographically or ecologically derived groups as a framework to which the botanical composition of archaeobotanical samples is compared. To do this, it is usually necessary to convert the ‘raw’ archaeobotanical counts into groupings of taxa based on relevant physical or ecological characteristics.

For example, weed seed characteristics such as size, weight and tendency to remain in ‘heads’ are relevant to an investigation of crop processing based on groups of ethnographically collected crop (and weed) samples derived from different stages in the crop processing sequence (Jones 1984, 1987).

A similar approach can be used to identify past crop husbandry remains on the basis of known associations of particular ecological characteristics of weed species with particular husbandry practices (Bogaard 2004).

These more goal-oriented approaches to the interpretation of archaeobotanical data are a useful complement to simple pattern searching which, in turn, can throw up unexpected patterns in the data.

References

• Bogaard, A. 2004. Neolithic farming in Central Europe : an archaeobotanical study of crop husbandry practices. London: Routledge
• Colledge, S. 1998. 'Identifying Pre-domestication Cultivation Using Multivariate Analysis,' in A. B. Damania, J. Valkoun, G. Willcox, and C. O. Qualset (eds.) The Origins of Agriculture and crop Domestication. Aleppo: ICARDA. pp. 121 - 131
• Jones, G. 1984. Interpretation of archaeological plant remains: ethnographic models from Greece, in W. van Zeist and W. A. Casparie (eds.) Plants and Ancient Man: Studies in Palaeoethnobotany. Rotterdam: 43-61.
• Jones, G. 1987 . A statistical approach to the archaeological identification of crop processing. Journal of Archaeological Science 14: 311-23.
• Jones, Charles, Bogaard and Hodgson (2010) 'Crops and weeds: the role of weed functional ecology in the identification of crop husbandry methods.' Journal of Archaeological Science 37. 70-77.Download