. . .

DISCUSSION

The points raised in the foregoing discussion can be summarized as a number of key issues which, we believe, those engaged in lead isotope studies directed towards archaeological provenancing will wish to address. These are concerned with the selection of both samples and of analytical data for source field characterization, and the determination of statistical confidence limits associated with the fields determined.

There is a need for a clear explanation of what type of material may be used to characterize ore source fields. Arguments have been presented that some types of material, particularly slags, that were used in this way in the past, are at risk of carrying a mixed lead isotope signature. Either these materials can or can not be used for ore source characterization. The practice of selecting for use only those measurements which conform to preconceived distributions (Sayre et al. 1992a) is unsound and should be abandoned. Similarly, although there has been much discussion of deviant data, particularly by Sayre et al. (1 992a), no justification has been presented for their exclusion from source field characterization other than their failure to fall within notional limits for the source field. The circularity of such arguments is inescapable. Whilst they are clearly aware that the inclusion of such data would broaden ore source fields, limiting the ability to attribute artefacts to specific sources, there has been a marked reluctance on the part of some workers to accept that ore source fields may truly lack the resolution that archaeometallurgists would like. In this context, it is interesting to contrast the cautious approach of German researchers (Pernicka et al. 1990) with the optimistic claims of Sayre et al. (1992a) and Gale and Stos-Gale (1992).

Another area of concern relates to the methods by which 90% confidence ellipses are added to ore source data on lead isotope ratio scatter plots. Clarification of the procedures used in their determination is necessary since, in many cases, points within the ellipse of 90% confidence are associated with two-sigma error bars which extend far beyond it. This creates a misleading impression of the separation of source fields and of the associations of analysed artefacts. A measurement may lie within the 90% confidence ellipse of one field whilst maintaining an appreciably high probability of association with another.

In view of these points and those raised elsewhere (Pernicka 1992; Reedy and Reedy 1992), it is with some trepidation that one reads the suggestion of Sayre et al. (1992a) that no matter how small a group of specimens that have been analysed from a given source there are always some worthwhile conclusions that can be drawn from them. After examining the data presented by them we are unable to share their optimism as to the possibility of using the source fields they have defined for questions regarding the provenance of artefacts from the eastern Mediterranean. In a region where many of the base metal ore bodies are apparently of similar age and difficult to define uniquely, false assignments of artefacts to specific ore deposits appear to be likely. On current evidence it would appear that overlap of the lead isotope ratios of eastern Mediterranean ore sources cannot, in reality, be reduced by the methods outlined by Sayre et al. (1992a). We take the view that aspects of the problem of ore source field characterization in the eastern Mediterranean require closer scrutiny and more detailed explanation before such procedures can be adopted. Additionally, any discussion of the evaluation of lead isotope data for archaeological provenancing would undoubtedly benefit from more detailed description of archaeologically relevant ore bodies and of archaeometric lead isotope methodology in the mainstream isotope geochemical literature.