Abstract
This paper discusses the use of "legacy data" in research on Roman Period iron smelting in the territory of the Przeworsk Culture in Magna Germania. The dataset includes results of 240 analyses of smelting slag and iron ores chemistry. A majority of these analyses were conducted in the 1950s-1980s. The quality of these data is far below present-day standards. Only major elements were reported, analytical methods were often not specified (although optical emission spectroscopy and wet chemical analyses can be assumed in such cases) and information on detection limits, precision and accuracy of the results is missing. In spite of this, a Principal Component Analysis-Agglomerative Hierarchical Clustering treatment successfully isolated observations from the three main iron smelting regions of the Przeworsk Culture (the Holy Cross Mountains, Masovia and Silesia). These results to a degree confirm a theory proposed in the 1960s by J. Piaskowski, according to whom it was possible to distinguish iron produced in the Holy Cross Mountains from the iron produced elsewhere in the territory of what is now Poland on the basis of metal characteristics. These findings will pave the way to the ongoing research project on the Przeworsk Culture metallurgy. It is also argued that, apart from a search for new methods in iron provenance studies, more attention should be paid to results of earlier analyses and to a compilation of legacy databases. The other result is an open and flexible Agglomerative Hierarchical Clustering R code to examine discrimination between production areas and to propose artefact provenance patterns in a convenient interactive way using the R development environment.