Abstract.
Good, scientific provenancing of lithic, ceramic or metal artefacts is achieved by the matching of the material constituents of that artefact to their correct geographical source(s). It depends on an accurate and precise description of these, often very altered, constituents and the ability to identify their precise, natural, unaltered, raw material precursors.
Most scientific provenancing is undertaken using petrographical and/or geochemical techniques and these, or better still, combinations of them, can be highly successful. However, the chances of success depend upon both geological factors inherent in the raw materials, including the extent of their outcrops and anthropogenic factors, notably the degree of preparation and physical and chemical alteration of the raw materials that take place during the artefact's manufacture.
These factors are reviewed as a background and explanation for the proposal that a central resource be established to bring together Welsh raw materials, well-contexted artefacts and all relevant literature and physical analyses. This assemblage should form the basis for creating a dedicated, geoarchaeological database, incorporating existing, acceptable, geochemical and petrographical data but mainly comprising new data. The next couple of decades are the last in which this can be done.

Introduction
Whilst it is true that no provenancing data are better than incorrect data (especially once they are in the literature) sound provenancing has much to contribute to archaeology. It is therefore discouraging that at a time when available methods for scientific provenancing have become more sophisticated, discriminatory and less destructive there is a growing distrust of the concept and an unwillingness in both the archaeological and geological communities to commit appropriate mental and financial resources to it.
Already it is, and increasingly it will become, difficult to obtain high quality, scientific provenancing data, particularly if these data are requested for a limited number of artefacts. Indeed, for single, poorly contexted artefacts the request is likely to be denied. There are a number of reasons for this reluctance.
"The days of provenancing archaeological artifacts with a hand lens and a garden full of rocks are over - or should be". (Ixer 1994, 10). The archaeological literature has accepted and repeats many examples of very precise provenancing based on poor quality or insufficient data obtained solely from hand specimen identification. Hence expectations, especially in terms of geographical exactitude and speed of reply, have become unrealistic, as are the beliefs that macroscopic identification is sufficiently diagnostic to obviate the need for further petrography and that common lithologies are simple to provenance (when the exact opposites are true).
Too often the request to identify and provenance artefacts (especially lithics) appears to have a poorly defined purpose, to be an afterthought or perhaps sought for the appearance of completeness. Little or no explanation of the context or significance of the artefacts, or what is expected from the provenance data, is given. As a consequence, data that are too detailed or inappropriate and of little apparent use to the archaeologist are produced and destined to be divorced from any main text only appearing within an appendix or archive. Little that is useful has been achieved and no one is satisfied.
Too often insufficient funding is offered. For a quick hand specimen identification funding is not necessary, but it is precisely these identifications or rather their imprecision that are the root cause of many errors. The difference between 'this is' and 'this looks like' or between 'this is a Lower Carboniferous sandstone with many petrographical similarities with sandstones cropping out at' and 'this is a sandstone, the nearest sandstones are at' is the difference between science and guesswork. Traditionally universities and museums have given considered identifications and provenancing information at no charge to the enquirer. This is no longer possible for, at the very least, all preparation time and costs have to be accounted for. Indeed, a considered identification is time consuming and deserves an appropriate reward.

Scientific constraints on provenancing.
Scientific provenancing as defined here is the application of field and laboratory geological techniques to help provide as precise a geographical source as possible for an artefact or for the raw materials used in that artefact's manufacture. The standard methods for the identification of geological raw materials (the necessary first steps towards scientific provenancing) are:

a) Petrographical, the visual identification of minerals and textures. The majority of lithics and ceramics can be identified visually using a hand lens or petrographically using a transmitted light microscope. Within the last decade the latter method has been supplemented by reflected light petrography to give 'total petrography' (Ixer 1994) where all the mineralogy can be investigated. Fine-grained materials, where petrography is difficult, are identified by X-ray diffraction techniques or scanning electron microscopy, ores by reflected light petrography and metals by metallography.

b) Geochemical, the chemical characterisation of material. All artefacts can be chemically characterised and the most common method is to determine the concentrations of selected trace elements. In the case of metals, trace element geochemistry may be augmented by stable isotope geochemistry, especially that of lead (Faure 1986).

Strangely, other than for fossiliferous rocks, geological dating of raw materials using their radiometric or palaeomagnetic properties has not been used very often as a provenancing tool in British archaeology. Exceptions, showing how effective this sort of dating can be, include Kelley et al. (1994) using potassium-argon dating for lithics and Budd et al (1999) using lead isotope dating of ores to provenance very early metal.

The likely success of scientific provenancing depends upon a number of factors.

A). Geological. The nature and natural distribution of the raw materials.
The petrographically distinctive, glaucophane-bearing schists found within a very limited number of small outcrops beneath the Marquis of Anglesey's Monument at Llanfairpwllgwyngyll close to the Menai Straits are unique in Wales and indeed in Britain. The presence of clasts of this schist within the non-plastic component of Bronze Age pottery from Anglesey is therefore clear proof of local manufacture for these pots (Williams and Jenkins 1997). In this rather extreme example provenancing was very precise and would have been rapid, simple and cheap, perhaps a few tens of pounds, as it was based on a pre-existing, adequate geological database.
Altered dolerites are common and widely distributed in Wales with large outcrops, but many have sufficiently distinctive petrographical and geochemical properties that they can be distinguished from each other. The best-known example of Welsh lithic provenancing is of dolerites, namely the Preseli Spotted Dolerites that comprise many of the Stonehenge Bluestones. By using some petrography and trace element geochemistry, including extending an existing geochemical database, Thorpe et al (1991) were able to match (provenance) individual bluestones to a limited number of outcrops at Preseli. Later, Ixer (1996; 1997), using very detailed petrography, in both transmitted and reflected light, moderated by the geochemical work of Thorpe et al (1991), felt he was able to refine further the match, so identifying individual bluestones with specific outcrops. In both cases the workers were able to establish the geological provenance of the stones, namely where the rocks were formed. Provenancing in this example was straightforward but time consuming and quite costly, in the region of high hundreds to low thousands of pounds. The majority of the costs were spent in extending the geological database into a geoarchaelogical one dedicated to the provenancing problem.
Somewhat controversially Thorpe et al (1991) argued that the Stonehenge monolithics were exploited from glacial erratics on Salisbury Plain rather than taken directly from their Preseli outcrops, so suggesting that for the bluestones their archaeological and geological provenances are different. This distinction between the two sorts of provenance is as important as it is contentious and the current inability to be able to recognise the difference is a serious problem in lithic studies. In the case of the exploitation of primary, in situ raw materials the geological and archaeological provenances are identical to each other and to the geographical location of the resource. However, for naturally transported materials (secondary sources), be they gold grains from a gold placer deposit, flint and chert from Recent river or marine gravels, or lithics from glacially transported boulder clay the geological and archaeological sources have become separated. The archaeological provenance (the site of exploitation) has been moved from the geological provenance (original outcrop), sometimes, as in the case of gold grains, huge distances. It is probable that the significance of secondary sources has been undervalued in provenance studies, for example the role of glacial erratics as the raw materials for polished stone axes.
Although Briggs (1989) and more recently Williams-Thorpe et al (1999a) have argued that erratics could be a viable source for some polished stone axes most workers are happier with the concept of dissemination of axes from discrete factory sites.
Finally many archaeologically important raw materials are not very chemically or petrographically approachable. Common and/or fine-grained lithologies like shale, siltstone, sandstone, altered volcanics but notably flint, chert, marbles, clay and quartz sands have little to offer routine geochemistry or petrography. Provenancing of these materials requires extensive fieldwork and complete sampling married to very detailed geochemistry and is, and will remain, very time consuming. The cost is almost open-ended with studies costing tens, and even hundreds of thousands of pounds, most of this cost is required to establish suitably sensitive techniques and to create the geoarchaeological data base.

B). Secondly the amount and degree of preparation needed to make the artefact.
Since most preparation methods distance an artefact from its raw materials in terms of chemistry and physical properties there is usually an inverse relationship between the degree of preparation needed to produce an artefact and the chances of its successful provenancing.
For mono-component artefacts like axe-heads or building blocks that have only undergone physical fashioning and perhaps a little heat treatment, there is a high chance of successful provenancing given a suitable lithology.
Multi-component, complex materials like mortars or clay-based artefacts, including ceramics, refractories, brick and tiles comprise clay ± a non-plastic component. The non-plastic component is most often lithic but may be organic (dung, straw, bone) or even earlier pot and the presence of more than one non-plastic component is possible. The artefact therefore combines two or more raw material signatures that may themselves have been modified prior to mixing - for example, the clay may be levigated, the non-plastic component, size-graded or burnt. Finally the clay-non-clay mixture is heat-treated so that many of its components undergo further mineralogical and geochemical modification.
Metals are, of course, the most difficult. They comprise just one, albeit the most valuable, end product of a number of such products including slags, ashes, gases, burnt clays etc. These in turn are formed at the end of a complex process or usually more than one process involving the mixing and radical heating of beneficiated ore(s), fluxes, fuels, pre-existing metals plus furnace linings, components that are themselves often highly processed. It is little wonder that successful metal provenancing (matching metal to ore) is contentious and, once metal recycling becomes significant, probably impossible (Budd et al 1995, 1996).

Developments in lithic provenancing and its future
Provenance studies on polished stone axes, including many from Wales, undertaken over the last fifty years by the Implement Petrology Committee (now the Implement Petrology Group) have been based almost exclusively on transmitted light petrography (Clough and Cummins 1979; 1988; Davis 1997). These studies have progressively demonstrated one of the main problems in lithic provenancing methodology namely 'petrographic drift'. In the case of the IPC this has resulted in the broadening of the definition of each axe-group leading to a less, rather than a more, secure provenance. Initially when the axe-groups were being determined and their provenance sought there was a strong internal consistency amongst the small group of petrographers leading to a set of 'tight' definitions but as the group of workers changed and expanded so this cohesion has progressively become eroded. Indeed in order to counter this erosion a number of the axe-groups have been and/or are being redefined and reassessed using combinations of petrographical and non-petrographical techniques (Markham 1997; Markham and Floyd 1998; Williams-Thorpe et al. 1999b; Jones and Williams-Thorpe 2001).
These reassessments have been part of a long-term investigation of the strengths and limitations of various petrographical, geophysical and geochemical techniques or combinations of them in lithic provenancing (Williams-Thorpe and Thorpe 1993; Stillman 1996; Potts et al 1997; Mandal 1997; Williams-Thorpe et al 1999b). A major finding of these studies has been to show that a combination of total petrography plus geochemistry (using either portable or non-portable XRF techniques) is very effective in provenancing lithics often to highly defined areas (Ixer 1994; 1997). The most recent work (Ixer et al in press) shows that it needed three petrographers (using transmitted and reflected light) plus a geochemist (using portable XRF) to provenance twelve 'random' axe-heads and illustrates the degree of effort, expertise and cost that is required. It shows the limitations of a single geologist/petrographer and that further advances, including rectifying the mistakes of the past, require the use of more than one independent determinative technique, team work including a number of specialised petrographers and a dedicated geoarchaeological database.
It is these last two requirements that are the problem. Finding a competent petrographer who is willing to undertake archaeological work is becoming increasingly hard and teams will be even more difficult to create. There are few all round jobbing petrographers and even fewer who are competent in both transmitted and reflected light and by their very nature these workers have exchanged breadth for much of the specialised knowledge of their colleagues. Sadly the last 30 years have seen a progressive decline in the amount of time given to the teaching and practice of both transmitted and reflected light petrography. There are many reasons for this decline, the skill is perceived to be old fashioned, time consuming, difficult to learn, dull to teach and unpopular with undergraduates. The use of the electron microprobe or scanning electron microscope in mineral identification has become a routine alternative to optical petrography so weakening the need for the skill. In addition most detailed petrography is learned 'on the job', often as the first step in an igneous/sedimentary/ore petrology investigation and as such only a sub-set of petrography is learned although often in great detail. Hence since most classically trained petrographers are now in the middle or towards the end of their careers they will be difficult to replace when dead or retired. It is therefore important to do as much archaeologically related petrography as is possible in the next 10-15 years whilst teams may still be created from appropriately qualified petrographers.
There is also the need to produce archaeologically dedicated geological databases. Despite there being a huge petrographical and geochemical database built up over the last hundred years or so, many rocks have yet to be fully characterised. Research on common rock types including many fine-grained or altered lithologies has been neglected in favour of 'exotic' and often coarse-grained rocks. The present day, detailed geographical location of a rock, its outcrop (geological provenance) and/or its natural dispersion pattern is of less significance to most geologists than is an understanding of the origin and subsequent history of the rock, its petrology. Hence fine-scale petrographical and geochemical variations in a rock, those over tens or hundreds of metres, may be of little importance in geology and so will not have been recorded or certainly not in sufficient detail to be of use in providing an acceptable archaeological provenance. These data need to be collected.
There is a similar and related problem that arises when a single source has provided more than one raw material as for example the historical extraction of coal, ironstone and clay from a single quarry. Here multiple materials share the same provenance but if they were used to manufacture very different sorts of artefacts their common association may not be recognised. Within the geological literature, in the unlikely event that these associated raw materials and their country rocks have been characterised in sufficient detail to be archaeologically useful it is doubtful that these data would be kept together. Therefore new data need to be collected or reassembled from the geological literature.

Developments in the provenancing of ores.
The problem of recognising and characterising multiple potential raw materials from a single locality is at its most acute at metalliferous mine sites and is compounded by the problem of correctly distinguishing between ores and non-ores.
Metal mining in Wales has been important since the Bronze Age, indeed, except for tin, the Principality has been a source for all the major metals during the last few millennia - sometimes, as at Parys Mountain, in the 1780s, even dominating the global supply and price.
Many metalliferous deposits are polymetallic and mineralogically zoned and so are potential sources for a number of different metals or sources for a number of different ores of the same metal, these changes reflecting and responding to changes in technology, metal use or metal value. As a consequence vast number of (Welsh) metal mines of all sizes have been mined on more than one occasion by more than one method and sometimes for more than one metal. For mining historians and even more so for prehistorians, there is the difficulty of determining for any given time what have been exploited, in other words in recognising the ore.
It is an irony that on many mine sites waste rock and sub-economic ores are plentiful, but true ore is difficult to find. This problem of ore recognition is at its greatest where there is the possibility that trace amounts of precious metals (especially gold) were being exploited. It has been an inability or unwillingness to discriminate correctly between ores and non-ores that has been one of the main reasons for the prevalence of the so-called Magpie School of Ore Provenance as defined by Ixer (1999). This inappropriate sampling strategy and the errors that have come from it have been major contributory factors in the present scepticism surrounding the effectiveness of lead isotope provenancing of metal and ores from Northwest Europe. Included in the scepticism are results from the Bronze Age Welsh copper mines at Parys Mountain and the Great Orme as well as a number of mine sites from southwest Ireland (Budd et al 1996; Ixer 2000; Ixer 2001).
It is therefore imperative that, prior to any attempt to match artefact metal to a mine site, especially if this is going to be done geochemically, ores are distinguished from by-product ores and from non-ore mineralised material. Ixer (1999) proposed a method for doing this. He has suggested that prior to any analytical work all samples should be subjected to ore-triage a method of classifying mineralised rock samples collected from a mine site into ore(s), by-product ore(s) and non-ore(s) based on metal grade, tonnage and mining and beneficiation characteristics. Table 1 shows the results of ore triage on the copper (and lead) mineralisation found at the Great Orme Llandudno (Ixer 2001). It illustrates how many and what types of data can be obtained, the degree of petrographical detail that is required to obtain these data, and how those data can be used to make decisions about what should be sampled for future work. As ore triage methodology is simple, based on field skills and ore petrography and requires little sample preparation it is relatively cheap. Its costs are perhaps one tenth of those of a few stable isotope analyses and the data that are produced are far less ambiguous and arguably more wide ranging than the geochemical data. Other Welsh mine sites, notably Parys Mountain and the polymetallic mines of Central Wales including Cwmystwyth, which have at least nine separate episodes of mineralisation (J. Mason pers. comm.), should be assessed by this method of ore triage prior to further characterisation.

Recommendations for Welsh Provenance Studies.
There is a need to establish a single, centralized resource dedicated to archaeologically and historically important Welsh raw materials, so creating a single definitive reference collection. The National Museum of Wales is the obvious place, as it has the geological, mineralogical/petrographical and archaeological expertise and many of the required facilities and in some cases already has the basis of the reference collection, as for example its collection of Welsh metalliferous ores.
The resource should comprise all relevant archaeological, historical and geological literature and a reference collection of well-characterised raw materials, to include lithified and unconsolidated rocks (clays, sands and gravels) and ores. Alongside these it should have representative artefacts collected from well-contexted archaeological/historical contexts.

1. Known and likely Welsh archaeologically significant raw materials should be identified and catalogued.
2. Representative geological material from identified raw material sites should be collected for petrographical and geochemical characterisation. Sampling, although comprehensive, should reflect the archaeological importance of the raw materials as well as being a record of the total geological variation at the site.
3. Resample artefacts/ores especially those with an established and scientific provenance, ensuring there is enough material for a polished thin section, a cut surface and for geochemical analysis. Redescribe, and if necessary correct, previous descriptions and augment published petrography with geochemistry and vice versa.
4. Create a dedicated geoarchaeological database, to include all published descriptions and to supplement these with additional petrographical and geochemical data.
5. Restrict the data sources. Routine geochemical data should be obtained by a single method, namely X-ray fluorescence or portable X-ray fluorescence obtained from a single laboratory. Petrography should be undertaken by the same team of petrographers.
6. This should be done in the next twenty years.

References

Briggs,S. 1989. 'Axe making traditions in Cumbrian stone'. Archaeological Journal. 146, 1-43.

Budd, P., Haggerty, R., Ixer, R.A., Scaife, B. and Thomas, R.G. 1999. 'Copper deposits in south-west England identified as a source of Copper Age metalwork'. http://www.archaeotrace.co.uk

Budd, P., Haggerty, R., Pollard, A.M., Scaife, B. and Thomas, R.G. 1996. 'Rethinking the quest for provenance'. Antiquity. 70, 168-174.

Budd, P., Pollard A.M., Scaife B. and Thomas R.G. 1995. Oxhide ingots, recycling and the Mediterranean metals trade. Journal of Mediterranean Archaeology 8(1), 1-32.

Clough, T.H. and Cummins, W.A. (eds). 1979. Stone Axe Studies. CBA Research Report. 23. London: Council for British Archaeology.

Clough, T.H. and Cummins, W.A. (eds). 1988. Stone Axe Studies. Volume 2. CBA Research Report. 67. London: Council for British Archaeology.

Davis, R.V. 1997. 'Implement petrology: 50 years of service to the archaeological community'. In A. Sinclair, E. Slater & J. Gowlett (eds), Archaeological Sciences 1995. Oxford: Oxbow Monograph, 64, 1-4.

Faure,G. 1986. Principles of Isotope Geology. New York John Wylie and Sons.

Ixer, R.A. 1994. 'Does ore petrography have a practical role in the finger-printing of rocks?' In N. Ashton and A. David (eds), Stories in Stone. 10-23. London: Lithics Studies Society Occasional Paper 4.

Ixer, R.A. 1996. 'Ore petrography and archaeological provenance'. Mineralogical Society Bulletin. 113, 17-19.

Ixer, R.A. 1997. 'Detailed provenancing of the Stonehenge dolerites using reflected light petrography: a return to the light'. In A. Sinclair, E. Slater & J. Gowlett (eds), Archaeological Sciences 1995, 11-17. Oxford: Oxbow Monograph, 64.

Ixer, R.A. 1999. 'The Role of Ore Geology and Ores in the Archaeological Provenancing of Metals'. In S.M.Young, M.Pollard, P. Budd & R.A.Ixer (eds), Metals in Antiquity, 43-52. British Archaeological Reports. 792.

Ixer, R.A. 2000. Further lead isotope data from Ross Island, Ireland. Historical Metallurgy Society Newsletter 43, 4-5.

Ixer R. A. 2001. 'An assessment of copper mineralization from the Great Orme Mine, Llandudno, North Wales, as ore in the Bronze Age'. Proceedings of the Yorkshire Geological Society. 53, 213-219.

Ixer, R.A., Williams -Thorpe, O., Bevins, R.E. and Chambers, A.D. In press 'A comparison between 'total petrography' and geochemistry using a portable XRF as provenancing tools for some Midlands axe-heads'. In Lithic Studies in the Year 2000. Natural History Museum of Wales.

Jones, M.C. and Williams-Thorpe, O. 2001 'An illustration of the use of an atypicality index in provenancing British stone axes'. Archaeometry 43, 1-16.

Kelley, S., Williams-Thorpe, O and Thorpe, R.S. 1994. 'Laser argon dating and geological provenancing of a stone axe from the Stonehenge environs'. Archaeometry. 36, 209-216.

Mandal, S. 1997. 'Striking the balance: the roles of petrology and geochemistry in stone axe studies in Ireland'. Archaeometry. 39, 289-308.

Markham, M. 1997. 'Geology and archaeology: a search for the source rock used by British Neolithic axe makers'. Open University Geological Society 25th Anniversary Edition. 18, 48-57.

Markham, M. and Floyd P.A. 1998. 'Geological fingerprinting of West Cornish greenstones as an aid to provenancing Neolithic axes'. Proceeding of the Ussher Society. 218-223.

Potts, P. J., Webb, P. C. & Williams-Thorpe, O. 1997. 'Investigation of a correction procedure for surface irregularity effects based on scatter peak intensities in the field analysis of geological and archaeological rock samples by portable X-ray fluorescence spectrometry'. Journal of Analytical Atomic Spectroscopy. 12, 769-776.

Stillman, C.J. 1996. 'Using geochemistry in archaeology'. Mineralogical Society Bulletin. 113, 11-15.

Thorpe, R.S., Williams-Thorpe, O., Jenkins, D.G., Watson, J.S. with Ixer, R.A. and Thomas, R.G. 1991. 'The geological sources and transport of the Bluestones of Stonehenge, Wiltshire, U.K'. Proceedings of the Prehistoric Society. 57, 103-157.

Williams, J. and Jenkins, D. 1997. 'Provenancing of prehistoric and Roman pottery using petrographic and mineralogical analysis: case studies from North Wales'. Archaeological Sciences 1995. 1-4. Oxford: Oxbow Monograph, 64, 92-99.

Williams-Thorpe, O. and Thorpe, R.S. 1993. 'Magnetic susceptibility used in non-destructive provenancing of Roman granite columns'. Archaeometry. 35, 185-195.

Williams-Thorpe, O., Aldiss, D., Rigby I.J. and Thorpe, R.S. 1999a. 'Geochemical provenancing of igneous glacial erratics from Southern Britain and implications for prehistoric stone implement provenancing'. Geoarchaeology. 14, 209-246.

Williams-Thorpe, O., Potts, P. J. & Webb, P. C. 1999b. 'Field-portable non-destructive analysis of lithic archaeological samples by X-ray fluorescence instrumentation using a mercury iodide detector: comparison with wavelength-dispersive XRF and a case study in British stone axe provenancing'. Journal of Archaeological Science. 26, 215-237.