Geometric Morphometrics and a 3D Caddo Vessel from the Sam Kaufman Site at SMU

While yes, this post contains a great interactive figure of an incredible Caddo vessel from the Sam Kaufman site (curated at Southern Methodist University #SMU), I wanted to take a moment to bring us full-circle, and return to the reason that I am scanning all of these vessels. Some time ago, I began getting fascinated by the high degree of variability in Caddo vessel shape and size, and have since also noticed some degree of variation in asymmetry.

At the same time that my interest in this topic began to increase, I also quickly realized just how much room for growth that there was in terms of applying geometric morphometrics to archaeological problems. It took time to get through the literature; and I kept finding new articles and book chapters listed in the references of my readings that had escaped the reach of my initial literature review.

Since I had been tinkering with social networking, I reached out to a colleague to assist me in building a citation network. That network has since been completed, and proved to be an invaluable asset in terms of not only centralizing the archaeological literature associated with geometric morphometrics, but it helped me to identify those works in the geometric morphometric literature that are most often cited (InDegree) and most important (identified using Google’s PageRank algorithm) to the overall network (click here to view the interactive network).

GMNET
Directed bipartite citation network for studies of geometric morphometrics in archaeology.

Having made my way through the literature, I began to think through the various methods of landmark and semilandmark applications that were available (based on the numerous research questions that could be asked of these data), and quickly found that the configurations were really only limited by my 3D modeling abilities. At that point, I ventured up to Lakewood, Colorado to spend some time working with the crew at Geomagic, where I learned how to use and employ the various features of Geomagic Design X and Control X (more here). Using these tools, I was able to devise a method of applying landmarks in a replicable manner using reference geometry that was built around the 3D mesh of the ceramic vessel. This first iteration of the landmark and semilandmark configuration worked very well for addressing some of my initial questions (see that in the video below).

From here, things began to get more complex. In October of last year, I headed to Portugal for what would be an important transitional period for my work; learning how to write scripts for, and run the various analyses in, the geomorph package in R. This opened something of a Pandora’s box for me, from which it is very likely that I will never fully recover. Needless to say, shape, form, allometry and asymmetry shifted quickly from a peripheral interest to something of a primary project.

SSP_2015_WSMProcrustesSuperimposition
Procrustes superimposition of landmarks and sliding adaptive semilandmarks used in the analysis of Caddo ceramics.

Since then, there have been dozens of iterations and developments in my landmark and semi-landmark configurations and analyses; many of which (particularly those associated with my morphological integration inquiries) are actively evolving. I was fortunate to receive funding for much of this work through the National Center for Preservation Technology and Training (NCPTT); a group that I have really enjoyed collaborating with.

As you interact with this vessel from the Sam Kaufman site, consider where (and how) you would apply landmarks/semi-landmarks to the 3D surface.

In terms of theory, analyses of geometric morphometrics can be couched within several interesting anthropological lines of inquiry that include, but are not limited to, (1) identifying the locus of a specific innovation, (2) the spatial and temporal dynamics of morphological variation for specific elements (neck, body, base, etc.) of ceramic design, (3) identifying or refining social networks used by specific Caddo polities/groups during temporal periods previously defined—primarily—through design-based seriations, (4) intra/inter-polity/group variation of shape, form, allometry and asymmetry for ceramic design, (5) potential trade relationships based upon the presence of a specific shape/form of vessel outside of known (assumed) social boundaries, and (6) the power or influence that shifted among and between polities through time. These considerations could be woven into discussions of communities of practice, craft specialization, ceramic technological organization, politics, religion, and—possibly—inter/intra-polity disputes and warfare. Furthermore, this research design has the capacity to inform greatly upon the evolution of ceramic design as it relates to the shape, form, allometry and asymmetry that occurs in Caddo vessels, and by adding the related qualitative measures to our results we might just have the potential to bolster evidence for human behaviors associated with ceramic production and use within the larger ancestral Caddo territory.

fighistcaddonet
Historic era (1680+) Caddo networks based upon ceramic types (as I tip my hat to Dr. Timothy K. Perttula). Incorporating elements of vessel shape, form, allometry and asymmetry may help us to further refine this current iteration of the Historic Caddo network.

Initially a development in the biological sciences, the study of geometric morphometrics in archaeology will no doubt include some interesting discussions regarding the various analytical and theoretical components that are most appropriate for a cultural system versus a biological system as we continue to press forward. There remains plenty of thinking left to do on this subject, but based upon the preliminary results, the capacity for geometric morphometrics to inform upon issues related to material culture and cultural systems could be enormous.

asymmetry
The very humble beginnings of exploring directional and fluctuating asymmetry in Caddo vessels.

Beyond the realm of empirical geometric morphometric studies, lies the domain of theoretical morphospace. Biologists have gainfully used theoretical morphospace to aid in clarifying issues of morphological change through time. They have done so by aggregating the results of geometric morphometric studies–a dialogue which would seemingly fit very well within the scope of anthropological and archaeological inquiry. By definition, theoretical morphospace represents the full range of possible morphologies for a group of artifacts; allowing investigators to posit, and contemplate, more- and less-adaptive morphologies (similarities and differences). It is within discussions of theoretical morphological transitions where I see the greatest promise for geometric morphometrics in archaeology; an ambit of inquiry in which the skeleton trees and topological properties of the artifacts tell us a much more dynamic story with regard to the progression of a particular shape (bottle, bowl, olla, etc.) through time. Within the context of my own long-term research design, theoretical morphospace seemingly holds much promise, and may represent the approach needed to identify, unlock and unpack a ceramic morphological transition that remains hidden in the various vessel shapes once employed by Caddo potters.

So, while the 3D images are fun to interact with–and have any number of preservation, access and outreach perks–my intention is to use them to bolster our discussions of shape, form, allometry and asymmetry in Caddo ceramic studies, and to use that evidence to posit a number of novel insights into the highly variable and dynamic prehistoric landscape that the Caddo people once commanded.

Many thanks to the Caddo Nation of Oklahoma and Southern Methodist University (Dr. Sunday Eiselt in particular) for the requisite permissions and access needed to scan this important artifact from the ancestral Caddo territory, and to present the 3D model here in color.

selden3dtagline300

Feature by 3D Systems

Selden_2015_3DS1

Selden_2015_3DS2

Selden_2015_3DS3

 To read this and other Success Stories from 3D Systems, click here. To view the original release of this industry publication, click here.

Selden3DTagline

3D Geometric Morphometrics of Projectile Points – Populating Splines

 

As we continue to think through the various spline configurations (see previous post here), I have been constructing a number of models based primarily on the efforts of previous analyses for both 2D and 3D geometric morphometrics. While this is representative merely of a (very) humble beginning, experimenting with the reconstruction of configurations used in other analyses could help us to better understand how we might begin to move toward a replicable consensus configuration (certainly some great examples out there).

ZSpline_1

In this example, I am using the framework of two splines that were created as a 3D mesh sketch in Design X (above). I then added 10 equidistant sections (below) between the top and bottom of the projectile point. While those would most likely be cut where they intersect with the splines above prior to populating point data, I wanted to see where the–equidistant–points would populate along the various profiles.

ZSpline_4

In this case, it may work best to cut each of the 10 sections where they intersect either the exterior or interior profiles (or both), prior to populating the LM/sLM data. I did not populate the 10th section (very near the point of the projectile) for this example, simply because it was almost impossible to view the location of Point 2 (which is defined by the confluence of the vector and the poly-vertices of the mesh at the tip of the projectile) when that particular section was populated.

ZSpline_2

Another question that we will need to ask ourselves sooner than later, is where do we reach a point of saturation or diminishing returns with regard to the number of LM/sLM data points? There is still much left to think about, and we will continue to move toward the definition of a replicable consensus configuration as we work through replicating (as closely as possible) the numerous configurations that have been used in the past.

Our tactics differ from many of our colleagues, due primarily to our efforts to devise a configuration aimed at performing an initial “sort” at the assemblage level–different configurations would then be used for each of the identified categories based on more specific attributes. In the coming months, I hope to share some of our minor successes, and–no doubt–numerous failures as we continue to work toward a suitable configuration.

As always, your comments and constructive criticisms are welcome. You can comment by clicking on Leave a Comment below, or you can email me directly at selden3d@gmail.com. 

Selden_Clovis


Selden3DTagline

Toward a Replicable LM/sLM Configuration for Projectile Points

Among the many topics that warrant further attention in 3D geometric morphometric studies of archaeological artifacts is the development of a replicable method for applying landmark (LM)/semi-landmark (sLM) data points to projectile points. While I use one of our 3D Clovis points to demonstrate our progress thus far, my interest lies more entrenched in the development of a consensus configuration for dart points in general. Extending our capacity to scan and analyze diagnostic points at the level of the assemblage by employing the same LM/sLM configuration is particularly attractive. While it is likely that this approach–much like that of ceramics–would result in a hierarchically-nested method of analysis, this could potentially yield a platform that minimizes subjectivity and semantics in classifying morphological variation. Importantly, LM/sLM configurations should be constructed to address specific research questions.

Selden_2015_RawWe begin by importing the 3D file into Design X (above). Note that the point is not aligned. The consensus configuration begins with one primary assumption upon which everything else is built – the vector. The vector is inserted along the principal axis of the point, and is defined by an algorithm. We then place a reference point at the confluence of the vector and the poly-vertices of the 3D mesh at the base of the projectile. A plane is then inserted at the juncture of the reference vector, point, and poly-vertices of the mesh. This configuration of reference geometry is then used to align the mesh.

Selden_2015_AlignedWe use the aligned mesh (above) to create a mesh sketch of the point’s profile, then extrude a surface around the 3D mesh. Deviations are subsequently calculated between the surface and the mesh, identifying the single point on the projectile that lies farthest from the central vector (below). We then insert a second plane along this widest profile of the point.

Selden_2015_VecDevOnce inserted, this widest planar surface (see image at top)–along with the various reference geometry created thus far–can be employed to create a seemingly unlimited variety of LM/sLM configurations. As long as those configurations remain based upon the reference geometry, replication of the geometric elements should be possible. The geometry provides the framework upon which the splines and other vectors can be created, then populated with data points.

 Selden_2015_P1Selden_2015_MSExtending our analyses beyond typical orthogonal measurements (length, width, thickness, stem width, etc.) can help us to better characterize the dynamic nature of projectile point morphology. While there is much work left to do on this front, we have begun to test various LM/sLM configurations (see one of these below).

Selden_2015_Potential

We will keep you posted as we make progress, and welcome any/all feedback as we continue to design and experiment with the wide range of LM/sLM configurations.

Selden_Clovis

robert z selden jr, archaeology, archeology, geometric morphometric, ceramic, analysis, mathematics, statistics, 3d, 3d scan

On Missing Data: 3D Morphometrics of Ceramic Artifacts

In the sample of complete and reconstructed Caddo NAGPRA vessels from the Turner Collection, many were found to include missing data (most often from sherds that were never recovered). While we have not been scanning vessels with large amounts of missing data–must be very close to complete–we needed to test the various methods by which those missing data can be reconstructed. Further, we wanted to explore the deviation of the results from the original mesh.

To do this, we used a whole/intact vessel from the Ellis Collection, cut a hole in the mesh, then used one of three functions in Geomagic Design X (defeature, fill holes, and edit boundaries) to generate new data over that area. Shifting over to Geomagic Verify, we use the original mesh as the nominal data, and the scan with missing data as the scan data to calculate the deviation between the two.

EditBoundaries_D1313

Results from the edit boundaries function.

FillHoles_D1313

Results from the fill holes function.

Defeature_D1313

Results from the defeature function.

In this case, the defeature function resulted in the lowest deviation from the original surface; however, this is not always the case. Each of the three functions was found to be successful in addressing missing data, and all warrant exploration on areas of the vessel that are geometrically similar to that where the missing data occurs to identify which function works best in each individual case. Additionally, the results of these comparisons should augment any publication as supplementary data.

My work with geometric morphometrics employs landmarks and sliding adaptive semilandmarks along a spline to compare various aspects of vessel shape, and selecting the correct function to address missing data in a sample could potentially impact those results. Through making an informed decision regarding which function to implement, we are mitigating a–potentially–higher degree of error within our sample.

selden3dtagline300