The aim of the research reported here was to establish the petrological and geochemical signatures of the Carrock Fell Gabbro; and to apply these to identify any implements, from a selection of 17, with provenance from this source (Table 6). The research was successful on two levels: firstly, the aims were fulfilled and seven implements (Table 7) were demonstrated to show secure Carrock Fell Gabbro provenance; and secondly, the research has highlighted important limitations to this approach and the potential for further research.
Petrological data. Generally, the optical data was good enough; it clearly demonstrated characteristic petrological textures and features in both the Carrock Fell rocks and the implement samples. Extensive alteration reduced the available information in some specimens. Only ten implements, of the 17 geochemically analysed, had available thin-sections and one implement remained unsectioned. This limited the proportion of implements investigated that can be demonstrated to show a secure provenance.
Sampling bias. Two implements showed a good geochemical affinity, but only moderate petrological similarities. While these implements do show Carrock Fell provenance, however, the actual lithotype utilised was not sampled in the field. This highlights a potential need to perform more extensive sampling of the Carrock Fell Complex as a whole, perhaps also including dyke rocks, xenolithic material, and the diabase (or dolerite) associated with the granophyre.
The PXRF dataset. The research has provided an independent assessment of analytical bias and error in the PXRF dataset. Statistical linear regression was used to estimate bias in the PXRF Carrock Fell rock data, which used a specimen 'surface window' of 25mm on a freshly cut flat surface as an estimate of the bulk composition. The data were found to be good enough, and only minor corrections for Ti were required. This is consistent with the findings of Williams-Thorpe et al. (1999a; 1999b), that analytical precision is good for data obtained from freshly cut flat surfaces. Of greater concern is the reliability of the unassessed errors in the PXRF implement dataset due to acquisition on weathered, curved and pitted implement surfaces. Williams-Thorpe et al. (1999a) have shown that while PXRF provides a valuable non-destructive analytical technique to investigate stone implements, analysis on weathered implement surfaces can yield results significantly different from the actual bulk composition, especially for K, Ca, Ti and Fe. Although bias resulting from curved and irregular surfaces can be corrected, error due to weathered surfaces is more difficult to assess and varies depending on rock type and weathering history (Williams-Thorpe et al. 1999b).
Reliability of element discriminators. The quality of interpretation drawn from geochemical analysis is significantly reduced if reliable discriminatory elements cannot be confidently established. Implements Car 8-1967 and Hu 300-1942-81 showed a poor to moderate geochemical similarity to the Carrock Fell Gabbro, although petrologically they bore little resemblance. In contrast, implements Car 14-1899-6 and Car A11 223 showed poor geochemical similarities, although the petrological similarities were very good. This could indicate a weakness in the methodology used, which is likely to be due to the inclusion of non-reliable elemental discriminants in the analysis. Williams-Thorpe et al. (1999a) have suggested that Zr, Sr and Ba are good discriminators for weathered surfaces, and that Sr, Zr and Y form good discriminators for the provenancing of fine volcanic implements (Meighan et al. 1993). Therefore, it is suggested that while the statistical methodology developed in their paper is sound, statistical results are highly dependent on the quality of data used.
PXRF sampled elements. By definition, gabbros are geochemically similar in regard to their major components, because they are grouped by similar mineral assemblages. Therefore, it is expected that most gabbros will show similar elemental concentrations of major elements, and trace elements must be used to discriminate between gabbros. PXRF analysis cannot measure Si (ppm), because it is too light to be captured by the instrument (Williams-Thorpe et al. 1999a). The concentration of Si would be useful in this research to plot the varying amounts of interstitial quartz and granophyric intergrowths. Similarly, it was not always possible to sample a wide range of rare earth elements using PXRF analysis, which may be a better diagnostic tool to discriminate between gabbroic rocks obtained from different geological sources.
Statistical methodology. Overall, the statistical model developed is sound enough and has yielded meaningful results. There is some evidence of geochemical exclusion of overall similar implements due to high sensitivity of the test, especially using PXRF data at the 1 SD level. Reliable discriminator elements must be established to obtain reliable results. Diagnostic trace elements including the more immobile rare earth element assemblages must be used in order to make confident interpretations. Larger sample sizes would significantly reduce apparent inconsistencies.