Is pXRF Now an Accepted 'Tried and True' Method in Archaeology?

By Mark Golitko, Associate Editor for Lithic Studies and Network Analysis

It goes without saying that new methods are often met with initial skepticism in most fields, and archaeometry is no exception. New techniques and methods often follow a predictable pattern, first being touted as the solution to all problems, followed by debate and exploration of the limits of the technique. Some techniques than find an established niche within the realm of things they are useful for, while others are rejected entirely (Collar et al. 2015). Seriation, for instance, has a long history going back to C.J. Thomsen’s three-age system, although in its developed form it is most associated with Flinders Petrie’s work at Diospolis Parva (Egypt). Despite the wide application of seriation in archaeology as the foundational means of developing chronologies, it was not until much later that empirical studies demonstrated that the popularity of styles did in fact follow something like the “battleship curves” that Petrie proposed (Renfrew and Bahn, 2019). The advent of radiocarbon dating in the late 1940s however demonstrated that while seriation might be good at establishing relative chronologies, it failed badly at establishing absolute anchored chronologies, and existing chronological frameworks were sometimes off by millennia (Renfrew 1973). 

Similarly, despite initial belief that radiocarbon dating could easily establish absolute chronologies, skepticism quickly set in as to whether radiocarbon dating could really provide the anchored chronologies it promised, and some archaeologists initially rejected the method in favor of retaining relative chronologies. It soon became evident that a host of factors beyond invariant radiogenic decay influenced the age of a sample recovered in archaeological context, including temporal variance in ¹⁴C production in the atmosphere, the old wood effect, reservoir effects, and so forth (Chapman and Wiley 2016). Even today, debate continues as to how best to link radiocarbon dates to past events—witness ongoing tension about the use of Bayesian modelling (Hamilton and Krus 2017).

Archaeometry has seen its own share of such debates over methodological validity, including heated disagreements over applications of lead isotope analysis for sourcing Mediterranean bronzes (Chapman and Wiley 2016), or the use of bulk chemistry to assign ceramics to production regions (Neff et al. 2006; Stoltman et al. 2006). As someone who entered the world of archaeological science around 2005, I witnessed firsthand as debates erupted over the application of portable X-ray fluorescence spectrometry (PXRF) (e.g., Frahm 2013; Speakman and Shackley 2013). Perhaps a low point of my career (in terms of personal morale in any case) was presenting my own first forays into PXRF and having a physicist in the audience stop me half way through my talk and spend the rest of my allotted time explaining what I had done wrong! They were however correct, I had not done the hard work to properly understand the data I had generated, something I spent a long time thereafter rectifying. Impressionistically, it seemed to me that at first, some practitioners distrusted on principal that PXRF could generate anything valid, part of a broader concern with wider applicability of scientific instrumentation and concern that many current users are insufficiently expert to ensure valid results (Killick 2015). This is of course a normal and predictable pattern, as noted. 

I recently attended the virtual Society for American Archaeology meeting (April 15th-17th) and the International Obsidian Conference (April 30th-May 2nd), at which numerous papers presented obsidian sourcing data collected by PXRF. Although I admit that I was not able to see every paper, nor engage with every conversation at these meetings, it seemed to me few issues were raised regarding methodology, which makes me wonder where PXRF is at this point in terms of acceptance. There still remains considerable debate over the right way to calibrate and report data, however, it seems that increasingly most practitioners have come to accept that both empirical calibrations (comparing energy peaks to measurements on reference materials) and so-called “Fundamental Parameters (FP)” methods (a variety of approaches based on using fundamental rules of physics) can produce valid data. However, one can’t just claim validity, it has to be demonstrated by showing how well a particular method works (e.g., Frahm 2017).

Despite arguments that empirical calibrations are preferable as they are “transparent,” while FP alogithms are “black boxes” (Speakman and Shackley 2013), I do wonder how much most practitioners actually understand of the data they generate using empirical calibrations. By far the most commonly utilized empirical calibration for obsidian in archaeology is that supplied with Bruker Tracer model PXRFs, which from the perspective of the user, simply outputs final calibrated concentration values, much as a FP algorithm would. In practice, these empirical methods may be just as much of a “black box” as values generated by proprietary FP methods. In other words, it is worth wondering how much of the distrust of PXRF has been dispelled by “experts” (i.e., analytical chemists and recognized archaeological scientists) taking back control of the method, and how much has been mitigated by actual increased awareness of how the instruments actually work by the average end user. 

Cumulative number of sourced obsidian pieces in Mesoamerica from 1972-2016 (Golitko 2019: Figure 3).

That said, I suspect that for obsidian, the most heated debates are in the past. Numerous papers are now published each year without there being much of an issue. In fact, in some world regions, the number of chemically sourced obsidian pieces is increasing exponentially at present, largely driven by the ease and low cost of using PXRF (Golitko 2019). For other materials, debates likely still lie in the future—there are for instance an increasing number of papers published using PXRF for non-destructive analysis of ceramics (e.g., Le Moine and Halperin 2021), but there remain issues about how best to process samples and calibrate results for these analyses, and whether PXRF is really an appropriate method for such analysis. That said, another common stage in the acceptance of any new method, as noted for radiocarbon dating, is recognizing the limitations of the technique—if PXRF was once presented as a sort of magic wand that could rapidly generate data on almost anything, the limitations of the technique have become more evident. This is of course no different than for any other method of chemical characterization. PXRF in this sense is no more problematic than INAA or SEM-EDS, so long as it is used in ways that suit its particular strengths and limitations. 

 

References

Chapman, R. and A. Wylie. 2016. Evidential reasoning in archaeology. London, Bloomsbury Academic.

Collar, A., F. Coward, T. Brughmans, and B.J. Mills. 2015. Networks in archaeology: phenomena, abstraction, representation. Journal of Archaeological Method and Theory 22: 1-32.

Frahm, E. 2013. Is obsidian sourcing about geochemistry or archaeology? A reply to Speakman and Shackley. Journal of Archaeological Science 40: 1444-1448.

Frahm, E. 2017. First hands-on tests of an Olympus Vanta Portable XRF analyzer to source Armenian obsidian artifacts. IAOS Bulletin 58: 8-23.

Golitko, M. 2019. The potential of obsidian “big data.” UISPP Journal 2(1): 83-98.

Hamilton, W.D., and A.M. Krus. 2018. The myths and realities of Bayesian chronological modeling revealed. American Antiquity 83(2): 187-203.

Killick, D. 2015. The awkward adolescence of archaeological science. Journal of Archaeological Science 56: 242-247.

LeMoine, J.-B., and C.T. Halperin. 2021. Comparing INAA and pXRF analytical methods for ceramics: a case study with Classic Maya wares. Journal of Archaeological Science: Reports 36: 102819.

Neff, H., J. Blomster, M.D. Glascock, R.L. Bishop, M.J. Blackman, M.D. Coe, G.L. Cowgill, A. Cyphers, R.A. Diehl, S. Houston, A.A. Joyce, C.P. Lipo, and M. Winter. Smokescreens in the provenance investigations of Early Formative Mesoamerican ceramics. Latin American Antiquity 17(1): 104-118.

Renfrew, C. 1973. Before Civilization. London, Jonathan Cape.

Renfrew, C., and P. Bahn. 2019. Archaeology: theories, methods, and practice, eighth edition. London, Thames & Hudson, Ltd.

Speakman, R.J., and M.S. Shackley. 2013. Silo science and portable XRF in archaeology: a response to Frahm. Journal of Archaeological Science 40: 1435-1443.

Stoltman, J.B., J. Marcus, K.V. Flannery, J.H. Burton, and R.G. Moyle. 2006. Petrographic evidence shows that pottery exchange between the Olmec and their neighbors was two-way. PNAS 102(32): 11213-11218.