Contents lists available at ScienceDirect Journal of Archaeological Science: Reports journal homepage: www.elsevier.com/locate/jasrep Using Fordisc software to assign obsidian artifacts to geological sources: Proof of concept Marin A. Pilloud ⁎ , Derek J. Reaux, Geoffrey M. Smith, Kristina M. Wiggins Department of Anthropology, University of Nevada, Reno, 1664 No. Virginia Street/MS0096, Reno, NV 89557, USA ARTICLE INFO Keywords: Multivariate statistics Great Basin Provenance studies X-ray fluorescence ABSTRACT In recent years, source provenance studies employing portable X-ray fluorescence (pXRF) technology have become commonplace in archaeology; however, they are not without critiques. Concerns center on the capability of instruments to produce valid results and researchers' abilities to accurately interpret those results and make correct source assignments. In this paper, we focus on the latter issue with a look towards statistical means of assigning artifacts to obsidian types using data provided by pXRF spectrometers. Using a sample of 677 obsidian artifacts from the northwestern Great Basin, we evaluate the ability of various approaches (principal components, cluster, and discriminant function analyses) to correctly assign artifacts to particular obsidian types. These multivariate methods generally work well to separate artifacts into different groups (i.e., obsidian types); however, they are less well-suited to assign individual artifacts to an obsidian source or type. We therefore tested the ability of the statistical program Fordisc, commonly used in forensic anthropology, to assign individual artifacts to specific geochemical obsidian sources or types. Our results indicate that Fordisc made accurate source assignments. Furthermore, because Fordisc provides probability values for different possible matches, it offers an advantage over other methods. 1. Introduction Applications of portable X-ray fluorescence (pXRF) technology to address archaeological questions have increased dramatically in recent years and many academic institutions and cultural resource manage- ment (CRM) firms now possess units. Although relatively expensive to purchase, pXRF instruments offer numerous benefits: (1) a non- destructive method to determine trace elements; (2) the ability to conduct in-field analyses (important when artifacts may not be collected); (3) the ability to characterize large numbers of artifacts in a relatively short amount of time; and (4) the elimination of commercial lab fees (Shackley, 2011, 2012). Their rapid and widespread adoption by researchers lacking previous experience in geochemical character- ization techniques has led some experienced analysts to express concern that some applications of pXRF technology have “no real foundation in science” (Shackley, 2012:2). Such concerns primarily center on issues related to repeatability (agreement between measurements collected under identical conditions at different times), reproducibility (agreement between measurements collected at different times under different conditions), accuracy (agreement between measurements collected using different instruments; for example, between pXRF and conven- tional wavelength-dispersive [WDXRF] and energy-dispersive [EDXRF] systems), and validity (the ability to collect and analyze data to differentiate raw material types and assign artifacts to those types) (Newlander et al., 2015). In this paper, we focus on the latter topic – validity – and how trace element data may be used to assign artifacts to geologic sources of raw material. We briefly review the range of approaches that analysts may use when making source assignments. We then present a novel method of data analysis that draws from the subfield of forensic anthropology. In this approach, we use the computer program Fordisc to assign artifacts to obsidian types. Analysts typically use Fordisc to help in establishing the biological profile for a set of unknown skeletal remains. To the best of our knowledge, our study represents the first time that Fordisc has been used in a source provenance study. Fordisc is easy to learn and use, provides custom-order discriminant functions, allows flexibility in analyses, and generates probabilities for individual group assignment as well as model performance. Although our sample of artifacts is small and we recognize potential limitations to the approach, using Fordisc to make source assignments represents an improvement over, or alternative to, other means of comparing univariate and bivariate trace element data. http://dx.doi.org/10.1016/j.jasrep.2017.04.023 Received 29 November 2016; Received in revised form 17 April 2017; Accepted 23 April 2017 ⁎ Corresponding author. E-mail addresses: mpilloud@unr.edu (M.A. Pilloud), dreaux@nevada.unr.edu (D.J. Reaux), geoffreys@unr.edu (G.M. Smith), krismw81@gmail.com (K.M. Wiggins). Journal of Archaeological Science: Reports 13 (2017) 428–434 2352-409X/ © 2017 Elsevier Ltd. All rights reserved. MARK