Since the mid-nineteenth century, artisans have produced masterful forgeries of Pre-Columbian art (Ekholm 1964; Kelker and Bruhns 2010). Forgeries can be found in museums around the world and sorting these out of museum collections is often difficult (Maclaren Walsh 2005). In this paper, we suggest that structured light scanning (SLS), when used with 3D point cloud processing software, can be an inexpensive, adaptable, and non-destructive technique for detecting certain types of forgeries. One common forgery technique is to use molds – both ancient and modern – to produce a suite of objects (Sellen 2004). Comparing detailed point cloud models of similar looking objects should be able to detect this type of forgery by determining if parts of the objects under study were made with the same mold. We tested this idea on three Zapotec ceramic effigies in a collection at the Royal Ontario Museum (ROM) in Toronto, Canada.
Ancient Zapotec culture, located in Southwest Mexico, produced complex ceramic effigy vessels from 200 BCE to 600 CE. The effigies, often referred to in the literature as “urns,” (Caso and Bernal 1952) were fabricated with the use of molds and hand modeling: mold-made central elements, such as a face or a headdress, were affixed to a larger structure that was usually a cylinder (Figure 1). Richly attired with the attributes of deities and other significant glyphic details, they are believed to represent venerated ancestors. Generally, urns have been found interred with the dead in tombs and shallow graves, and they vary greatly in size, ranging from 15 cm to over a meter high. They were often produced in series, repeating the same effigy four or five times over (Sellen 2007; Urcid 2013). Regrettably, the vast majority of urns in private and public collections lack a firmly established archaeological context and are the fruit of illicit excavation; a considerable number of urns in collections are also forgeries (Goedicke et al. 1992; Sellen 2004; Shaplin and Zimmerman 1978; Mongne 1987, 2000).
Early in the twentieth century, the demand for this type of pre-Hispanic ware increased, fomenting a cottage industry in fakes (Brulotte 2012). Difficult to distinguish from ancient wares, forgeries of Zapotec urns flooded the market and began to find their way into museum collections. The potters used the same techniques to produce ceramics as their forefathers had centuries before them, suggesting that the much maligned “fakers” of today can also be seen as part of what the art historian Mary Miller has called an “enduring tradition” (personal communication, 2021).
Since the 1970s, multiple studies have attempted to identify more recent ceramic creations and separate them from the ancient artifacts. This research is not only an important curatorial goal but is a crucial tool for scholars more generally who seek to study past material evidence without contamination from the present. Previous research on Zapotec urn forgeries classified them into groups and speculated on their origins (Mongne 1987, 2000; Shaplin and Zimmerman 1978), but few scholars asked how the forgeries were made. Posing this question reveals some of the methods used by artisans in the faking industry in Oaxaca. Howard Leigh, for example, was an artist and collector who witnessed clandestine workshops operating in the 1950s and 60s. He suggested that, “Fakers sometimes made molds for their little heads, but you usually can tell because there are no broken surfaces. They smooth that over and don’t reproduce the broken surfaces of the head. But they churn out these little heads” (Wren 1979: 9). Observations like this one not only provide insights into how forgeries are manufactured, but also can help explain how ancient artifacts could have been made.
There are two ways to detect forgeries: by the human eye – a visual analysis also known as connoisseurship – and by methods devised by modern science. Since the 1940s, more scientific methods of authentication have been valued over the connoisseur’s intuitive approach. Many of these methods, however, are time consuming, expensive, destructive, and require specialized knowledge and equipment, making them difficult to routinely employ in museum collections. A multi-method and holistic authentication approach tailored to the particular problems of each object can therefore be more successful, requiring expertise from disciplines in both the sciences and the humanities (Fleming 1975; La Niece 1997: 175; Berger et al. 2022). The identification of fake Zapotec urns in museum collections has benefitted from the development of this multi-method approach, combining a variety of archaeological science techniques with more in-depth studies on iconography and the history of collecting that can develop a more refined “eye” for authenticity (Jennings and Sellen 2018).
Since the 1960s, museums with Zapotec urns in their holdings have been suspected of harboring fakes. Howard Leigh, with his first-hand knowledge of the faking industry, was among the first to counter the opinion of well-known archaeologists and condemned a number of the objects they presented as fake (Boos 1964). In the mid-seventies, Philippa Shaplin, an art historian, estimated that perhaps 30% of the urns in older collections such as those of the Royal Ontario Museum, Toronto, Musée du Quai Branly, Paris, and the Ethnographic Museum in Berlin, were spurious. Determined to ferret out these creations with a more objective test, she teamed up with physicist David Zimmerman and used the technique of thermoluminescence (TL) dating to test urns from a large collection at the Saint Louis Art Museum. Their subsequent article not only identified the forgeries in the collection but also made a critical assessment of the stylistic criteria for determining authenticity (Shaplin and Zimmerman 1978). These same researchers applied a TL test to 36 urns at the Royal Ontario Museum and concluded that 34 of these were fakes (unpublished report, 28 December 1977). Following their lead over a decade later, Goedicke, Henshel and Wagner (1992) used TL and neutron activation analysis in an examination of Berlin’s extensive holding of Zapotec urns, identifying dozens of fakes with the first technique and pinpointing the provenience of the clay bodies with the second.
In 1999 the collection at the ROM was revisited by a group of researchers from the National Autonomous University of Mexico (UNAM). They examined 45 effigy vessels using TL and Proton Induced Particle Emission (PIXE), including some of the vessels previously studied by Shaplin and Zimmerman. The UNAM study not only confirmed these earlier results but demonstrated that the collection also had many ancient artefacts. In addition, the UNAM researchers supplied information regarding the origin of the ceramics through a historical investigation of the collection and iconographic analysis of the objects (Sellen et al. 2000). These combined approaches determined that Constantine Rickards, the English collector who sold these objects to the ROM in 1919, was behind many of the fakes produced in the early twentieth century. With the help of local artisans, he had employed molds taken from the ancient artifacts in his collection to produce pastiche fakes (Sellen 2004). Finally, in 2015, a pastiche object from the Rickards collection was analyzed using several different methods including provenance studies, X-radiography, computed tomography scan, thermoluminescence, X-ray fluorescence analysis, petrographic analysis, and pigment and adhesive analysis (Jennings and Sellen 2018).
The extensive study of the ROM’s collection demonstrates that even the most experienced connoisseurs will have problems authenticating Zapotec ceramics. To aid the “eye,” three-dimensional modeling provides an effective method for comparing mold-made objects. SLS, to our knowledge, has not yet been used to study art forgeries of this type. The closest parallel may be Gherardelli et al.’s 2014 discussion of laser scanning to produce virtual models from plaster piece molds of the Richard-Ginori porcelain factory. As a pilot study, we analyzed three effigy vessels in the ROM’s collection that had been deemed twentieth-century creations by virtue of a TL test carried out in 1977 (Figure 2). The faces on the vessels are similar and we wanted to determine if they were made using the same mold.
The three objects chosen for analyses are at the smaller end of the corpus of Zapotec effigy vessels, measuring 18 cm in height. They are kneeling female figures with their hands crossed over their chests. Each figure sports a different headdress that connects to lateral flanges with decoration. This type of seated posture for Zapotec effigies is common in the ancient canon, thus without the use of TL to determine their age it would be difficult to assess their authenticity based solely on connoisseurship. Employing high-definition 3D scan and point cloud processing software, we produced 3D scans of the entirety of the artefacts to obtain a better sense of the similarities and differences between urns in both their appearance and construction technique. We then focused our analysis on the faces of the effigy vessels to look for indications that the faces were made with a similar mold.
Structured light three-dimensional scanning measures the shape of an object using projected light patterns and a camera system. The scanning method works by projecting a sequence of narrow bands of light onto an object. The curvature of the object distorts the bands from the perspective of the light source, and the shape of the object can be rendered by a geometric reconstruction. Error ranges vary depending on the size of the objects, equipment used, and other factors, but can be as low as 10 micrometres (0.00039 in) (Chen et al. 2015).
The three forged Zapotec urns discussed in this article were scanned in the fall of 2020 using a HP 3D Structured Light Scanner Pro S3 that was operated using David laser scanning software. After calibration, the process began by first scanning, and then removing, the background around each urn. The urn was then scanned through a full 360° angle, with a total of 16 scans taken as it rotated on the turntable. After the scans were cleaned for unwanted noise around the object, the scans’ texture and surface features were used to complete a global fine registration to align all scans. To fill in small holes in the aligned scans, the “fill holes” function was run at 0.1% to create a fused scan: the three-dimensional model of the urn that could then be exported to other programs.
The scanning process was repeated until a three-dimensional model was generated for each of the three forged Zapotec urns. The models were visually compared to the actual urns to ensure accuracy. Photographs were also taken of each urn, focusing on the face where we suspected that the same mold was used across the three objects (Figures 3, 4, 5). The object file format (.obj) represents three-dimensional shapes as a series of vertices along with texture information to allow the surface colours to be displayed if required. Object files are widely used by 3D graphics application programs, including CloudCompare. In this case, only the surface morphology in the form of a mesh was used for comparative analysis.
CloudCompare is open-access software originally designed to manipulate and analyze large point clouds generated by terrestrial laser scanners (Girardeau-Montaut 2006). Initial applications of this software within the energy and geology industries have expanded to include forensic analysis (Fournier et al. 2019; Miranda et al. 2018), building conservation (Valero et al. 2018, 2019), and, more recently, archaeology (Poux et al. 2017). Apart from point clouds, CloudCompare is able to analyse triangular mesh data in a variety of file formats (.ply, .obj, .stl, .off, .fbx) that are common outputs of photogrammetric modelling or structured light scanning.
The three SLS models were of a sufficiently high resolution to allow both coarse-grained comparison between the vessels in their entirety and more fine-grained comparisons of parts of the urns. In this case, we chose to more closely study the face because of the similarities seen as a result of the more coarse-grained comparisons. For this fine-grained comparison, the .obj models of each vessel were inspected using MeshLab, an open-access 3D modelling software, to isolate the facial areas of interest from the rest of the model prior to importation into CloudCompare. With the same regions of each model extracted and saved in a .ply (Polygon Line format or Stanford Triangle Format) file format, the resulting mesh was imported into CloudCompare. The .ply and .obj files generated during the project are permanently stored at the Royal Ontario Museum and available upon request.
Three separate analyses were conducted to cover all possible comparative permutations: 1) 917.4.112 to 917.4.116, 2) 917.4.112 to 917.4.34 and 3) 917.4.116 to 917.4.34. Each of these comparative analyses were conducted in the same fashion, first by loading two meshes to CloudCompare and second, by automatically aligning the two meshes using the Fine Registration (ICP) tool. This step assumes that both meshes are already roughly registered in the same coordinate space, are of the similar scale and that they both represent variations of the same object. Once aligned, the absolute difference between each pair of meshes was calculated using the Cloud/Mesh (C2M) Distance tool.
The output of this comparison was represented in the combined models as a variation of colour across a designated spectrum or scalar field (Figures 6, 7, 8). The colour ramp and intensity of the scalar field of the cloud/mesh distance (C2M) output was then adjusted to better visualize the differences between the two meshes. In this case, a blue-green-yellow-red colour ramp was chosen, and the scalar field was adjusted to allow the centre region of the spectrum of green to yellow to represent areas that approach an absolute difference of 0 mm, or in other words, a 100% match. Variations towards the blue or red extremes of this chosen spectrum represent the highest difference between specific points or regions in each pair of meshes.
The same set of analyses was conducted as part of our fine-grained analysis of the faces (Figures 9, 10, 11). The same scalar field comparison was made between the meshes using a blue-green-yellow-red colour map. In addition, a histogram of each C2M distance calculation was also generated along with the models to provide a graphical representation of the number and degree of similarity between individual points in each mesh along a gradient of absolute difference in mm. Finally, a record of the basic statistical measures for each C2M difference calculation was generated for each comparison including minimum, maximum and average distances between points, the standard deviation, and a maximum error (see Figures 9, 10, 11).
The 3D models show how the Zapotec effigy vessels were made by a combination of molds and hand modeling. The method of manufacture demonstrates the techniques, both ancient and modern, used by potters in the Oaxaca region (Shaplin 1975). Using the scalar fields generated by CloudCompare allows one to identify areas of difference (seen in red) more quickly from areas of similarity (seen in green) and can visually illustrate differences between the urns in such features as the diameter of the cylinder and the angle of attachment of the headdress (see Figures 6, 7, 8). A close visual inspection, nonetheless, would provide the same results. The value of comparing point cloud models therefore comes in more detailed comparisons of parts of the vessel that are based on quantitative distinctions between vessels. In the more coarse-grained scalar field that includes the entirety of the effigy vessels, the faces look nearly identical. Does a more fine-grained analysis of the faces support the argument that all three faces were made using the same mold (see Figures 9, 10, 11)?
The scalar fields of the faces show very high degrees of similarity, with minimal deviation in the meshes across the majority of the surface of each pair of models. The most extreme variation (represented in red) is largely because the 917.4.34 model has sustained some damage to the tip of the nose, and as such registers as a significant difference between models in that region of the face. Other areas of relatively high deviation tend to occur towards the edges of the facial area, likely due to variation in the post-molding, pre-firing manipulation of each ceramic element as they were bonded and burnished. This signature of workmanship is also highlighted in the comparison of 917.4.112/917.4.116, as the bridge of the nose in the latter is visibly flattened (see Figure 7) and clearly registers as a significant variation in the CloudCompare analysis.
The scalar fields, when combined with the C2M distance calculations, makes possible a comparison between the effigy vessels that is impossible to obtain with even the most trained eye. Slight variations in the curvature of a cheek, for example, would be difficult to discern without CloudCompare, and these differences would neither be quantifiable nor replicable. At best, the faces might “feel different” to the connoisseur, but the differences would not be easy to articulate. By using CloudCompare to examine precise 3D models of objects, we can definitively identify those features that are shared between objects. In our case study, all three urns share the same face.
Do structured light scanning or other high-resolution 3D scanning methods add another technique to our arsenal for identifying forgeries? We have shown that the technique can help identify the use of the same mold across a group of artifacts. In the case of Zapotec urns, the presence of molds is not necessarily indicative of a forger’s fingerprint, given that ancient potters also used this technique. However, previous studies determined that the ancient ceramicist would most often employ molds to produce identical urns corresponding to a cosmogenic order that reflected the four directions, often including the center. Generally speaking, these sets were fabricated in a series of four or five effigies (Figure 12).
The forgers, on the other hand, relied on a model of production that produced a diversity of creations that could be considered plausible when compared to archaeological specimens; and a wide range of objects for sale meant greater financial return. Invariably these are pastiche creations. A potter applied different motifs copied from ancient works to a base figure in order to yield a diversity of desirable objects, a concept illustrated by the three urns that we chose for our study. Had the forgers used their molds to churn out hundreds of identical objects, their game would have been given up very early, because customers would wonder why their purchases looked the same as others. Eventually the strategy of creating pastiche wares based on originals also had an inherent weakness: the forgers’ ignorance of the ancient Zapotec aesthetic canon, especially relative to details such as glyphs, resulted in the assembling of iconographic vocabularies in an unrecognizable grammar.
We now have a face that we know was featured on three urns made in the early twentieth century. We can search for this face in the remaining effigy urns in our Zapotec collection and send the mesh file to other museums where curators would like to examine their own objects. If the face in our case study was taken from an ancient model, then a few of these vessels may yet prove to be pre-Columbian (Sellen 2004). At the very least, we have demonstrated that this face is part of a corpus of molded elements of Zapotec effigy vessels that have been found in forgeries. As our fund of knowledge of the ancient Zapotec aesthetic canon continues to grow – through iconographic scholarship and further scientific testing – discerning dissonance in this culture’s visual vocabulary is becoming increasingly refined. Techniques such as those described in this article, used as part of a multidisciplinary approach, advance our overall knowledge of the forger’s modus operandi and therefore constitute a useful tool in the endeavor to identify spurious objects from the ancient Zapotec and other cultures.
The authors have no competing interests to declare.
Berger, ME, Moreau, C and Lemaitre, S. 2022. Reconnecting Collections: Provenance, Material Analysis, and Iconographic Studies of Mesoamerican Turquoise Mosaics and Related Pieces. In: Martinez, MM, Sears, EL and Sieg, LE (eds.), Contextualizing Museum Collections at the Smithsonian Institution. The Relevance of Collection-Based Research in the Twenty-First Century. Washington, DC: Smithsonian Scholarly Press.
Boos, FH. 1964. Corpus Antiquitatum Americanensium. Vol. 3. Colecciones Leigh, Museo Frissell de arte zapoteca, Smithsonian Institution, y otras. Mexico City: Instituto Nacional de Antropología e Historia y Unión Académique Internationale.
Brulotte, RL. 2012. Between Art and Artifact: Archaeological Replicas and Cultural Productions in Oaxaca, Mexico. Austin: University of Texas Press. DOI: https://doi.org/10.7560/737792
Chen, F, Brown, GM, Song, M and Yoshizawa, T. 2015. Three-dimensional shape measurement. In: Yoshizawa, T (ed.), Handbook of Optical Metrology, 517–540. New York: CRC Press. DOI: https://doi.org/10.1201/b18328-20
Ekholm, GF. 1964. The problem of fakes in Pre-Columbian art. Curator, 7(1): 19–32. DOI: https://doi.org/10.1111/j.2151-6952.1964.tb01723.x
Fournier, G, Savall, F, Nasr, K, Telmon, N and Maret, D. 2019. Three-dimensional analysis of bitemarks using an intraoral scanner. Forensic science international, 301(August): 1–5. DOI: https://doi.org/10.1016/j.forsciint.2019.05.006
Gherardelli, M, Balleri, R, Di, S and Adembri, G. 2014. 3D laser scanning of historic molds for documenting the Richard-Ginori factory collection. Journal of the American Institute for Conservation, 53(3): 145–158. DOI: https://doi.org/10.1179/1945233014Y.0000000023
Maclaren Walsh, J. 2005. What is real? A new look at Pre-Columbian Mesoamerican collections. Anthronotes, 26(1): 1–7, 17–19. DOI: https://doi.org/10.5479/10088/22411
Miranda, GE, Wilkinson, C, Roughley, M, Leite Beaini, T and Haltenhoff Melani, RF. 2018. Assessment of accuracy and recognition of three-dimensional computerized forensic craniofacial reconstruction. PLoS One, 13(5): e0196770. DOI: https://doi.org/10.1371/journal.pone.0196770
Mongne, P. 1987. Les urnes funéraires Zapoteques: collectionnisme et contrefacon. Journal de la Société des Américanistes, 73: 7–49. DOI: https://doi.org/10.3406/jsa.1987.1022
Poux, F, Neuville, R, Van Wersch, L, Nys, GA and Billen, R. 2017. 3D point clouds in archaeology: Advances in acquisition, processing and knowledge integration applied to quasi-planar objects. Geosciences, 7(4): 96. DOI: https://doi.org/10.3390/geosciences7040096
Sellen, A. 2015. The Orphans of the Muse. Archaeological Collecting in Nineteenth-Century Oaxaca. Mérida: Centro Peninsular en Humanidades y Ciencias Sociales, Universidad Nacional Autónoma de Mexico.
Sellen, A, Litvak, J, Ruvalcaba, JL, Bucio, L, Lazos, L and Ortiz Díaz, E. 2000. How to spot a fake: Preliminary results for an authenticity study on Zapotec urns from the Rickards Collection at the Royal Ontario Museum. Proceedings of the 32nd International Symposium on Archaeometry. CD-ROM. Mexico City: Universidad Nacional Autónoma de Mexico.
Shaplin, PD and Zimmerman, D. 1978. Thermoluminescence and style in the authentication of ceramic sculpture from Oaxaca, Mexico. Archaeometry, 20(1): 47–54. DOI: https://doi.org/10.1111/j.1475-4754.1978.tb00211.x
Valero, E, Bosché, F and Forster, A. 2018. Automatic segmentation of 3D point clouds of rubble masonry walls, and its application to building surveying, repair and maintenance. Automation in Construction, 96: 29–39. DOI: https://doi.org/10.1016/j.autcon.2018.08.018
Valero, E, Bosché, F, Forster, A, M’Beirick, I, Wilson, L, Turmel, A and Hyslop, E. 2019. Development of a novel open tool for the segmentation of 3D point clouds of masonry walls. 14th International Congress on the Deterioration and Conservation of Stone. pp. 981–987. University of Göttingen: Göttingen.