For artifact interpretation, it is an enormous advantage to know from where an artifact or its raw material has come.  Information on the source of artifact material can be utilized to model trade networks, and questions about the procurement and use of the material by prehistoric societies can also be addressed.  One hope is to answer, if instance, whether a specific material was traded locally or over large distances.  The exchange mode by which the raw material or finished artifacts changed hands can be also studied.  Trade implies social contacts and is a possible source of cultural change as ideas are transmitted.  The value of the material can be explored as well, due to differences in the trade of commodities and prestige goods.  Matters of territory, access to resources, and technology may also be investigated.  The archaeological objective is, in the end, to be able to make inferences about economic organization, political structure, social stratification, and ideologies of the societies involved.  It is hoped that provenancing artifacts will allow archaeologists to make more meaningful inferences than those based simply on style.  In fact, prior to modern analytical techniques, stylistic traits were one of the only ways that a source could potentially be identified, in much the same way that they used to “date” artifacts before radiocarbon dating.  With chemical analysis, the source, or provenance, can be determined for a much wider range of archaeological materials.

    Provenance studies are based on patterns of elements, which vary in geological materials as a result of their different histories.  Chemical analysis of a particular deposit would then, in theory, reveal a characteristic pattern of elements.  Accordingly, artifacts manufactured of the material from this deposit should have the same pattern.  This characteristic pattern is often called a compositional “fingerprint.”  This refers, of course, to the unique patterns of ridges found on our fingertips.  No two individuals, not even identical twins, have identical fingerprints.  In several ways, a fingerprint is an apt nickname for a characteristic pattern of elements.  Our fingerprints are, of course, utilized as a means of identification, and compositional fingerprints are used as a means of identifying source materials.  Siblings usually have fingerprints with the same kind of ridge patterns, and artifacts with the same geologic “parentage” have similar compositional fingerprints.  Fingerprints are classified by pattern types, the size of those patterns, and their location on the finger.  Similarly, compositional fingerprints can be classified by the elements that are present, the amounts of the elements, and their distribution in a specimen.  Unfortunately, this analogy is not a perfect one.

    Another analogy, and maybe a more accurate one, for compositional fingerprints are ceramic typologies.  Ceramics are frequently used as indicators to date archaeological sites and components.  Archaeologists also utilize ceramics to study a variety of cultural aspects, such as settlement patterns and exchange networks.  They are sorted into typologies based on their technological (clay, temper, etc.), morphological (size, shape, etc.), and stylistic features (cord marking, painting, etc.).  Consider, for example, three ceramic typologies common in Minnesota: Oneota, Laurel, and Blackduck.  The attributes of Oneota ceramics are crushed shell temper, buff or buff-gray color, smooth surfaces, a height less than its width, a round shape with a rounded base, lip notches, and rectilinear decorations on the rim (McKusick 1964; Straffin 1971).  Laurel ceramics principally have thin walls, a smooth surface, and inscribed decorations made with a comb-like tool known as a dentate stamp (EMuseum 1999).  Blackduck ceramics are generally rounded with thick necks and rims.  Crushed rock is used as temper in most Blackduck vessels, and the colors vary from yellow to gray (Budak 1981).  Most characteristic of the Blackduck typology are its common decorations: punctates, cord impressions, brushing, and various decorations around the lip of the pot (Anfinson 1979).

    These typologies are artificial constructs, idealized classifications by archaeologists, utilized to sort ceramics by particular times and places.  Ceramic typologies are derived from a collection of sherds, and usually no one sherd possesses all properties of the class into which it has been sorted.  Of the three typologies defined above, some characteristics are distinctive, and others are shared by two or three.  It is the overall pattern of the traits that characterizes a typology, not any one.  Further, there are variations within a particular typology, such as the color of Blackduck ware and whether a Laurel pot has a smooth or dentate-stamped surface.  There are also, of course, very similar ceramic typologies, and sherds will be discovered that could be separated into more than one typology.  The same circumstances occur in compositional fingerprinting.  There exist variations within geological materials, and a deposit could have a fingerprint akin to that of other deposits.

    Therefore, such compositional fingerprints are, in essence, the same as other archaeological typologies.  Typologies are formulated to establish cultural affiliation, and their formation depends on the categorization of artifact attributes.  Trends in archaeological typologies are used to delineate cultural units and patterns.  Compositional fingerprints are no different, and they are used to source artifacts based not on form or decoration but composition.  Like ceramic typologies, the fingerprints are used to make inferences about the societies that fashioned the artifacts.  It must be remembered, though, that ceramic typologies answer questions about where artifacts were made while provenance studies involve the source of the raw material from which they were manufactured.

    Similarly, typologies intended to answer questions about where artifacts were manufactured have also been formulated based on chemical patterns.  These are not true “provenance studies” in our sense of the term -- they use analytical techniques to characterize and group similar artifacts, not trace them back to a geologic source.  Banterla et al. (1973), for example, utilized neutron activation analysis to classify terra sigillata potsherds “into homogeneous groups when it [was] impossible to place them stylistically or by any other means” (209).  Neutron activation analysis was also utilized by Barrandon et al. (1977) to sort Constantinian coins among different Roman mints.  Such studies can be certainly useful but are not “provenance studies” because a source of the raw material is not sought.  This is not to say, however, that the definition of a “source” is clear-cut.

    How a source should be characterized is also a subject of discussion and debate.  Glascock, in the same volume as Green (1998), contends that, in some provenance studies, “only a fraction of the information potentially available was recovered.  In other instances, poorly planned studies came to conclusions that were later refuted” (1998:20).  He lists a variety of problems: “failure to locate all possible sources, collection and analysis of too few specimens from each source,. . . analysis of samples from modern road cuts rather than prehistoric quarries, analysis of too-few elements, and failure to identify the most critical elements useful for discriminating between particular regions or individual sources” (20).  Also with respect to obsidian characterization, Shackley (1998) explains that “what [archaeologists] previously defined as a single source appears to be much more variable than originally conceived” (5).  He criticizes that it “is no longer enough to chemically characterize a source of obsidian by grabbing five samples from a road cut” and overlook its wider distribution in the region (6).  In their studies of native copper, Rapp et al. (1990) discovered that differences in sample size and collection methods affected the ‘tightness’ of trace-element fingerprints (236).  For instance, fingerprints for small deposits and small sections of large deposits were more defined than those of large native copper deposits and entire mines (Rapp et al. 2000:57).  Thus, the size chosen to define a “source” is an important consideration in provenance studies.

    Besides concerns about sampling and statistical tests, critics have also challenged the overall methodology of provenancing, that is, the so-called Provenance Postulate.  Glascock (2002) asserts:

It was not until the Provenance Postulate was proposed by Weigand et al. (1977:24) that an explicit statement of the main requirement for the success of sourcing by measurement of chemical differences was expressed.  More recently, Neff (2001:107-108) pointed out that the same requirements are necessary if one uses mineralogical or isotopic properties to source artifacts... he proposed a rewording of the Provenance Postulate: ‘Sourcing is possible as long as there exists some qualitative or quantitative chemical or mineralogical difference between natural sources that exceeds the qualitative or quantitative variation within each source’ (2).

Rapp et al. (2000) break down the Provenance Postulate into three components:

Provenance determinations have three major components: (1) locating and adequate sampling for analysis of all potential geologic source deposits for the specific artifact material in question; (2) using an analytic method that has the sensitivity and scope to provide diagnostic signatures for each geologic deposit as well as the artifacts in question; and (3) use of a statistical or data analysis technique that has the power to evaluate the data and then assign artifacts to source deposits when the data support it (4).

They point out that two intrinsic problems exist in this postulate.  The first, that all sources must be sufficiently sampled, has been discussed.  The second is that “it must be established that the artifact has not undergone any chemical or physical alteration that would invalidate direct comparison of the artifact with the same component material from known deposits” (2000:5).  There exist two parts to this issue: anthropogenic alteration and post-depositional alteration of material.  Even if humans did not change the composition of, for instance, clay or native copper, burial conditions might have done so.  This is a particular problem with archaeological ceramics.  For instance, Krywonos et al. (1982) reported that the concentration of sodium in Roman and Islamic coarse wares of western Cyrenaica and Crete was due to absorption of saline water in the region.  For metal artifacts, this effect is often limited to the surface and is usually known as either surface enrichment or, more accurately, surface depletion.  Condamin and Picon (1964,1965) concluded that, due to the post-deposition conditions, copper and silver can migrate within Roman denarii and produce a silver layer as thick as .2 mm on their surfaces.  Surface enrichment of archaeological metals is also described by Hall (1961), Banks and Hall (1963), Hall and Roberts (1962), Wyttenbach and Hermann (1966), Hawkes et al. (1966), and Meyers (1969), to name a few.  To counter such enrichment, Hall (1971) proposes that, when a surface- or spot-analytical technique is utilized, the surface of an artifact should be “rubbed down” to expose a “fresh” surface that is more representative of its interior.

    Some critics have even challenged the fundamental assumption that geologic material from a specific source exhibits a “significant and recognizable” elemental or isotopic pattern (Goad 1980: 17).  In other words, provenancing necessitates that each source has a distinctive fingerprint and that intra-source variations are less than the inter-source variations.  Inhomogeneities will always exist in geologic material, so this is a non-trivial issue.  Nevertheless, most researchers bypass it and simply presume that compositional delineations can be made among geographic regions.  A project that did thorough research in this issue was the University of Minnesota-Duluth Copper Project (Rapp et al. 1980, Rapp et al. 1984, Rapp et al. 1990a, Rapp et al. 1990b, Allert et al. 1991, Rapp et al. 2000).  A wide range of geologic variables that can influence composition: parent composition; environmental conditions such as mass, temperature, host rock structure, and cooling rate; later thermal events; and weathering, to name a few.  The UMD Copper Project researchers propose: “With such an array of geological processes and conditions which may affect ore occurrences, it would be strange indeed if the odds did not favor the development of chemically-unique individual deposits” (Rapp et al. 2000: 36).  The researchers, however, did more than conjecture that fingerprints would be unique; they set out to study the compositional variabilities within and among copper deposits.  Their work did find statistically significant differences; however, it was noted that inhomogeneities made “definitions of the sources less precise than one would like” (Rapp et al. 2000:3).

    The best answer concerning which elements and what statistical technique should be used is that researchers must first determine and then use “whatever works best.”  Although the process of science can involve deductive reasoning, the initial investigations in provenance studies are inductive, requiring observations to find any patterns first, before any hypothesis can be tested.  To continue with the example of the UMD Copper Project, Rapp and his colleagues had their samples analyzed, at one point, for as many as 46 different elements; however, only ten elements were utilized for their source characterizations and artifact assignments.  These elements were selected based on initial test results, geochemical tendencies, statistical constraints, and limitations of neutron activation analysis, which can detect less than two-thirds of the naturally occurring elements.  Consequently, Rapp et al. admit: “we do not want to imply that these are the only or even always the most useful elements for native copper sourcing” (2000:57).  Indeed, some other element that NAA cannot detect -- lead, for instance -- might have been suitable if it could have been measured.  In a perfect world, all elements could be detected with a single, low-cost analytical technique, and there could be unlimited access to supercomputers that run sophisticated statistical analyses.  So long as funding is an issue, however, archaeologists have to compromise and, as mentioned, use “whatever works best.”

    Related to the limitations of neutron activation analysis, which is a bulk analytical technique, is an issue of characterization.  Geochemical characterization asks three major questions: (1) What elements are present in the material?  (2) How much of the elements are present?  and  (3) Where are the elements?  Bulk techniques leave this last question unanswered because they cannot acquire any information about the spatial distributions of elements, trace or otherwise.  A helpful analogy is that of a lithic scatter.  Consider a circular lithic scatter in the middle of a field.  This debris scatter from prehistoric flaked-stone tool manufacture covers an area of 10 square meters, and it is comprised of about 2000 flakes of chert.  The field is one kilometer by one kilometer in size.  A typical one-meter by one-meter test unit in the middle of the scatter would indicate that it has a density of roughly 200 pieces of debris per square meter.  Other test units scattered across the field would not produce any debitage.  Such test units, systematically distributed, could yield a clear picture of the distribution of the debris.  If, though, the test unit were increased to the size of the full field, the result would differ; averaged over a square kilometer, the density would be .002 pieces of lithic debris per square meter. This result offers no information about the spatial distribution of the debitage within the field, and it even obscures the presence of a dense lithic scatter at the center of the field.  

    The difficulty is that few analytical techniques, namely particle beam methods, can determine the distributions of elements in a sample.  In 2001 and 2002, I utilized electron microprobe analysis, one such particle beam method, to obtain information about the micro-scale distribution of elements in native copper.  It was determined that one element was distributed uniformly while all others were distributed heterogeneously, sometimes as discrete inclusions.  These inclusions, while all primarily silver, differed in composition and morphology for the different sources.  Such studies of elements’ distributions are uncommon in provenancing.  Leute (1987) views microprobes as “one of the most useful instruments for characterization with respect to morphology and chemical composition.  This has long been recognized in general materials science, but is not less true for archaeometry” (124).  I would argue with Leute that electron microprobes have not been considered useful for provenance research, and as a result, information about the spatial distribution of trace elements is virtually non- existent.  The situation, however, is different for ancient technology studies, in which microscopy is commonly employed and micro-scale structures are important features.