You are on page 1of 27

Lithic Technology

ISSN: 0197-7261 (Print) 2051-6185 (Online) Journal homepage: http://www.tandfonline.com/loi/ylit20

Exploring New Approaches to Lithic Analysis: Laser


Scanning and Geometric Morphometrics

Michael J. Shott & Brian W. Trail

To cite this article: Michael J. Shott & Brian W. Trail (2010) Exploring New Approaches to Lithic
Analysis: Laser Scanning and Geometric Morphometrics, Lithic Technology, 35:2, 195-220, DOI:
10.1080/01977261.2010.11721090

To link to this article: http://dx.doi.org/10.1080/01977261.2010.11721090

Published online: 01 Apr 2016.

Submit your article to this journal

Article views: 20

View related articles

Citing articles: 6 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=ylit20

Download by: [George Mason University] Date: 13 June 2016, At: 21:19
EXPLORING NEW APPROACHES
TO LITHIC ANALYSIS:
LASER SCANNING AND
GEOMETRIC MORPHOMETRieS

Michael J. Shott
Downloaded by [George Mason University] at 21:19 13 June 2016

Brian W. Trail

ABSTRACT

However they measure stone tools, archaeologists typically use orthogonal-


mutually perpendicular-dimensions like length and width. Although useful,
orthogonal dimensions do not capture the geometric corifiguration ofspecimens.
We describe an alternative approach to lithic analysis that involves three-
dimensional laser scanning andgeometn"c morphometn"cs. 11zis approach preserves
much geometn"c informah"on and generally is morefaithfol to whole-obfectform.
Using landmarks-points at equivalent posih"ons on different specimens-it also
exploits poweiful analyh"cal techniques not traditionally used in lithic analysis. We
z1/ustrate the approach and its ana/yh"calpotential in the study ofa small collection
ofPaleoindianjlutedpoints.

"Palaeolithic archaeology has been slow to adopt methods that enable the comparative
morphometric analysis of lithic variability... " (Lycett 2009:79).

Michael J. Shott and Brian W. Trail, Department of Classical Studies, Anthropology and Archaeology, University of
Akron, Akron, OH, USA 44325-1910. shott@uakron.edu ; bwt7@uakron.edu
19§ Lithic Jec:bnologg, volume 35, 2
Laser scanning technology was developed with of theoretical perspectives and using a range of
industrial and medical applications in mind. But methods. Yet measurement of stone tools is a
it offers exciting prospects for archaeology as well, common practice that facilitates many approaches
not least in the analysis of stone tools. Lithic to lithic analysis. Measurement is intended to
analysts are just beginning to use laser-scanning characterize tools at once by size.and form. Lithic
technology and associated analytical techniques. analysts have developed many measurement
Befitting the growing range of applications, schemes that vary with the questions they ask of
instrumentation and software for laser scanning tools and assemblages.
span a wide range of technical specifications and
prices. As archaeologists begin to use it, naturally
their own applications occupy much of this range. ORTHOGONALI'IY
Yet, even now, comparatively few scanning
applications in archaeology have appeared in print. Despite their diversity, most measurement
schemes involve linear dimensions, usually
This brief report has two purposes. First, we orthogonal (i.e., mutually perpendicular) to one
review recent developments in lithic analysis that another and to Cartesian axes. At a minimum,
Downloaded by [George Mason University] at 21:19 13 June 2016

involve scanning technology and geometric these include length, width and thickness. There
morphometries. Second, we describe one variant is no doubting the value of standard measurement
of scanning technology and explore its potential for schemes, which capture basic information about
lithic analysis in the detailed morphometric specimen size and form. Yet orthogonal
characterization of stone tools. measurement schemes have limitations, or at least
are not ideal for all purposes.

THEORY AND METHOD IN Three shortcomings are particularly acute.


UTHIC ANALYSIS First, orthogonal dimensions reduce complex
wholes to relatively crude stick-figure caricatures;
As in any science, theory explains patteming in for instance, they reduce whole-tool length,
the variation found in lithic assemblages in complete with variation in other dimensions and
culturally or behaviorally meaningful terms. changes in outline form along the longitudinal axis,
Method serves theory by supplying the means to to a single scaled line. Second, orthogonal measu-
acquire the data whose variation is measured and rement preserves little information about the
explained. But method is not mere adjunct to spatial or geometric relationship between dimen-
theory such that, for instance, theory specifies the sions such as length, width and thickness. In
evidence that method then acquires and common practice, of course, these dimensions are
characterizes. Instead, method and theory exist in defined as mutually orthogonal; to that limited
dynamic tension, each influencing the other. As extent, orthogonal dimensions preserve geometric
methods change and improve, often independently information. But they preserve only direction, not
of theory, they can alter the way that data are location, because the positions of maximum width
characterized, thereby begging questions that or thickness relative to length are not specified.
available theory may not answer or even imagine. Instead, each dimension is treated as isolated or
A textbook example is the development of disembodied. While whole-object geometry can be
radiocarbon dating, a theoretical accomplishment approximated by, for instance, noting the position
in physics that became a valuable method in of maximum width along the longitudinal axis,
archaeology. It stimulated improved methods to approximations tend to be crude, limited models
collect datable organics and better chronometric of object form. Figure 1 shows the data reduction
control that in turn allowed archaeologists to revise and loss of geometric information typical of
interpretations of the past (e.g., Renfrew 1973). orthogonal coding schemes. Third, some funct-
ional characteristics of stone tools can be difficult
Lithic analysis addresses questions about past to capture using orthogonal dimensions. Tip acuity
cultures that their use of stone reveals. Because and cross-section (both form and area) are
stone is implicated in many aspects of ancient important characteristics in theory (Hughes 1998)
cultures' functioning, analysis necessarily and analysis (Shea 2005:824; Moncel etal2009)
addresses a wide range of questions from a range that linear dimensions capture poorly, if at all.
Shott and Trail- EJu>lorin& New Ap_proacbes to Uthlc Analysis: ... 197

1\
l \

n
\ I \

i
I
I
I '( =>
Downloaded by [George Mason University] at 21:19 13 June 2016

Figure 1. Data reduction and loss of geometric information in conventional orthogonal measurement. The left
pair of figures shows common orthogonal dimensions superimposed upon a point's plan and section.
The center pair shows the same dimensions without superimposed plans, indicating data reduction from
whole-object outline form to stick-figure orthogonal dimensions. The set of isolated orthogonal dimen-
sions (right) show the loss of geometric information.

Mindful of the limitations of orthogonal (2003) also used polar coordinates to


schemes, a few pioneering studies have explored reconceptualize a sequence of defined Illinois
alternatives. Holmer (1978:7-8) manually digitized Valley triangular-point types as a complex
two-dimensional coordinates of positions on the continuum of multivariate variation in a single
outline of Great Basin points to define morphotype. Instead of a sequence of successive
mathematical functions that distinguished types. appearance and decline of discrete formal types,
Gero and Mazzullo (1984) used Fourier series to the approach focused attention on the longer-term
model stone tool outlines. In a neglected classic rate and direction of change in that morphotype.
study that illustrated the typological implications In these and perhaps other studies, conventional
of pattern and extent of resharpening, Hoffman orthogonal dimensions would not have revealed the
(1985) characterized the outlines of southeastern patterns of variation found.
U.S. points by vectors of varying length and angle
emanating from a central position, using a polar- Whatever the virtues of these approaches, most
coordinate template. Wynn and Tierson (1990) remain variations on orthogonal themes. Until
used a similar approach to measure size and outline recently, it has been impractical to capture more
form of Acheulean bifaces. Henton and Durand geometric information without disproportionate
(1991) used raster methods to capture scaled investment of coding effort by use of dozens of
outlines of points and compared outlines for variables that inter-correlate to provide largely
symmetry and other purposes. And Morrow and redundant information. Whole-object form has
Morrow's (1999) hybrid approach used some proven much easier to appreciate than to measure.
orthogonal dimensions in a broader, morphometric
context.
GEOMETRIC MORPHOMETRICS
More recently, Bradbury and Carr (2003)
studied Early Archaic bifurcate-base types from the Beginning in the 1990s, biologists and biological
Tennessee Valley. They, too, usedoutlineformand anthropologists adopted geometric morphometric
polar coordinates, but also CAD technology, to (GM) approaches to the description and analysis
evaluate (and question) the formal integrity of of complex whole objects (e.g., O'Higgins andJones
several defined types and to distinguish functional 1998; Slice 2005b). "Morphometric" means
modes based in part on blade symmetry. Shott measurement and description simultaneously of
198 Lithic Technology, volume 35, 2
size and shape, capturing the spatial or geometric Arizona collection. Thulman (2006) analyzed
relationship between measured dimensions or Clovis and southeastern Suwanee points by
plotted points. GM data, in technical terms, "can calculating mostly standard orthogonal dimensions
be mapped back into physical space to achieve from digital images. Castifieira et al. (2007)
...visualizations that are frequently not possible analyzed outline form of Fluted Fishtail points from
with alternative methods" (Slice 2oosa:s). Put Uruguay and Argentina. Cardillo (2oo6)
differently, GM records more detailed information conducted a similar analysis on mid-Holocene
about whole-object form along dimensions and at points from Patagonia. Burnett and Otarola-
points that may not fall along common axes, and Castillo (2008) studied progressive modification
facilitates the treatment and analysis of such after retouch in a set of experimental Elko Corner-
detailed, geometric data. Notched specimens. Buchanan's (2006; Buchanan
and Collard 2007; Buchanan and Hamilton 2009)
First and most important among its virtues is is the most ambitious and detailed of recent 2D
that GM preserves and records inherently research, involving a wide-ranging North American
geometric information, viz. the relative positions sample of Clovis, Folsom and other bifaces. These
Downloaded by [George Mason University] at 21:19 13 June 2016

of the coded points, called "landmarks." Depending scholars defined a set of dimensions that links
on the breadth and detail oflandmark designation, landmark points at the tip, base corners and other
and unlike orthogonal approaches, GM preserves locations of specimens. Clearly, GM has
outline form, three-dimensional form, some considerable potential to redefine lithic analysis by
aspects of cross-section form, and other inherently replacing relatively few orthogonal dimensions
geometric information. By comparison, orthogonal with more, and more detailed, non-orthogonal
schemes record only the value of particular ones.
dimensions, preserving no information on the
geometric relationship between dimensions or Three-dimensional (3D) approaches are more
points along them (Tompkins 1993; Crompton recent. Riel-Salvatore et al (2002) explored the
2007). potential of 3D visualization for core-refitting.
Lycett et al (2006; see also Lycett 2009) used a
Besides its fundamental geometric character, variety of contact digitizers to capture landmark
GM possesses other virtues. It permits detailed points on Acheulian bifaces. Crompton (2007) and
description of object size and shape, and controls Moncel et al (2009) analyzed Middle Paleolithic
for size variation by scaling all objects under study tools considered projectile points, in part by 3D
to a constant size. Controlling for size is useful for characterization of tip form. Grosman et al (2008)
typological and other, but not all, purposes, because used 3D laser scanning to model Lower Paleolithic
"variation in artefact shape is usually associated handaxe form and to eliminate measurement error
with artefact size" (Crompton 2007:2). Ironically, from ambiguity in the orientation of specimens
its ability to control for size makes GM an especially relative to orthogonal axes. Sumner and Riddle
powerful means of allometric analysis. GM uses (2oo8) modeled Acheulian tools using
thin-plate splines for graphical expression of photogrammetry, not laser scanning, to produce
differences in forms between specimens or types. 3D digital models. These studies used 3D digital
For these reasons, GM is a promising analytical data as models of specimen size or form for the
approach to variation in three-dimensional size and kinds of analyses discussed at length below. In
form of stone tools. addition, Lin et al (2010) measured cortical surface
area on flakes, demonstrating the value of 3D
GM of stone tools is difficult without either digital models for other, equally valid, analytical
scanning or similar digitizing technology. purposes.
Accordingly, in recent years archaeologists have
begun to use scanning or digitizing and GM 2D models possess the virtue of smaller size and
methods to study form in a range of lithic types. simplicity and the vice of reducing 3D objects to
To date, most studies involve GM or similar 2D approximations. Their comparative economy
analysis of two-dimensional (2D) images. of data capture is a great practical advantage. Yet
Tompkins' (1993) neglected pioneering study 3D methods are progressing rapidly, and
involved a continental sample of Clovis-age points, comparative advantages in time, cost and model
many of them replica casts in the University of complexity diminish through time. What is more,
Sbott and TraU - Explorini New Approaches to Lithic Analysis: ... 199
2D and 3D variables do not always correlate anatomical features like cranial suture intersections or
strongly (e.g., Archer and Braun 2010:2070). To the tips of dorsal fins. Type .Dlandmarks are maxima
that extent, 3D methods and variables measured of cwves "associated with local structures usually with
using them are preferable because they are more biomechanical implications" (Slice 2005a:9). To
faithful to the shape and dimensionality of the O'Higgins and Jones, Type 11 landmarks may be
original objects. homologous in "geometric rather than compositional"
respects (1998:255). Typemlandrnarks are geometric
3D approaches include contact digitizing, which extremes or endpoints, for instance, the extreme point
captures only selected landmark points on of entire bodies, limbs or organs. They are "deficient"
specimens, so use of such data by other researchers (O'HigginsandJones2oo6;Slice2oosa:9)becausethey
is limited. They also include laser-scanning, which do not reduce to a precise anatomical point or because
produces complete, closed ("water-tight") digital "their displacement is meaningful principally in a single
models on which analysts may choose any number direction representing the length... of the defining
of analytical points which can change or increase segment" (Bookstein 1991:65), z:e., because they are
in number as analytical goals change. defined only in relation to other points.
Downloaded by [George Mason University] at 21:19 13 June 2016

Assimilating Bookstein landmark concepts to stone


GEOMEI'RICMORPHOMEI'RICS IN tools is inherently ambiguous (Lycett 2009:81). In one
THEORY AND PRACTICE sense, there can be no strict equivalents in stone to 1Ype
I and n landmarks, whim are defined explicitly with
GM arose in biology and paleobiology, so it owes its reference to organismal anatomy. Tools have no
founding concepts and terminology to those fields. Like "discrete configurations of tissue," so no strict 1Ype I
many approaches, GM has evolved a distinctive landmarks, nor are there "biomeclJ.anical" implications
vocabulary that must be mastered in order to ofcwve maxima, and thus no 1Ype IT landmarks. Yet if
communicate clearly. Hence this briefsemantic primer, landmarks definitions are broadened from their
whim follows Slice (2oos:a3-4). Slzapeis all geometric biological context, some degree of correspondence at
properties ofobjects that are unrelated to location, scale least is arguable.
and orientation. S)U>isproperties ofobjects that pertain
to distance or dimension. Fonn refers simultaneously Stone tool tips are points where opposing edges
to properties that involve both size and shape. Thus, converge, so undeniably are discrete points. On
"shape" and "form" have different meanings in GM. projectile points, tips also are essential functional
Shape is form without regard to size; alternatively, form characters. Similarly, base corners are discrete points
is shape scaled to size. Most biological GM analyses where curves intersect and have their own functional
concern shape alone. But size and form can be very significance, related to hafting. Therefore, tips and base
important in lithic analysis, not least in the study of corners might be considered Type I or Type 11
allometric resharpening of retouched tools (e.g., Iovita landmarks, as local points or "structures" that possess
2009; Shott et al. 2007). clear functional implications. Yet tips, base corners and
other landmark points discussed below also are
GM analyzes the geometric configuration of points, "geometricextremesorendpoints"likeBookstein's'IYpe
called landmarks, located on the surfaces of 2D or 3D m landmarks.
objects. There are two somewhat opposed views on the
nature and validity oflandmarks. One, associated with Semi-landmarks
Bookstein (1991), defined three types of landmarks.
Central to this approach is the concept of biological In the Bookstein view, landmark points are not
homology, shared ancestiy or derivation of anatomical adequate characterizations of all morphometricdata or
constructs between taxa. MacLeod(1999, 2008)tooka the corresponding properties of physical objects like
broader topological, not strict biological, view to artifacts; forms can include continuous sets of
homology. points that comprise curve outlines, surfaces or
areas. Accordingly, semi-landmarks are arbitrary
Bookstein points along edges, curves or other geometric forms
that collectively trace the forms but do not mark
Following Bookstein (1991:63-65; see also Slice meaningful specific points along them. Bookstein
20053:9), Type Ilandmarks are biologically discrete (1997:232) more or less formally defined semi-
200 Lithic Tecbnologu, volume 35, 2

landmarks as distinct from landmarks because, 2009). A partial solution is to analyze the same
although they share "some of the criteria that make specimens both with and without Type Ill
landmarks useful for scientific analysis," they landmarks and to compare results.
nevertheless "fail to be true landmarks ... [because]
they exist only in the context of a group average." MacLeod
In this somewhat cryptic passage, Bookstein
distinguished semi-landmarks from landmarks of MacLeod (1999, 2008) took a different
any type, including Type Ill ones. If Type Ill approach that largely moots distinctions between
landmarks denote extremes or end points, then Bookstein landmarks types. Defining landmarks
semi-landmarks are not equivalent to Type Ill as "nothing more or less than objectively defined
landmarks. Niewoehner (2005:289-290) used coordinate locations that bear some topological,
semi-landmarks to characterize irregularities on anatomical, functional, etc. correspondence across
metacarpal joint surfaces, Mitteroecker and all objects" (2008:154) under study, MacLeod
Bookstein (2008:946) to map cranial form. Their argued that biological homology pertained to entire
first use in stone-tool analysis occurred separately structures (e.g., organs, skeletal elements), not to
Downloaded by [George Mason University] at 21:19 13 June 2016

in Argentina (e.g., Cardillo 2006) and Britain (e.g., specific points within them. Accordingly, he
Lycett etal2oo6). questioned the validity of specific anatomical
locations as different landmark types, and
Whatever the conceptual and analytical validity advocated topological over biological homology. In
of Type 1-111 landmarks, in the Bookstein view practice, MacLeod also considered many Bookstein
semi-landmarks are not equivalent to them. They landmarks to be subjectively placed arbitrary
do not mark discrete points on the surfaces, edges points along complex curves. The thrust of
or endpoints of objects, but merely arbitrary points MacLeod's argument was to question the biological
along lines or curves. The arbitrary placement of validity ofBookstein landmarks and the sometimes
semi-landmarks can produce spurious pattern in labored distinctions between their various types
thin-plate splines, in the form of "bending energy" and, in the process, to reconcile landmark and
(Gunz et al 2005) that is the product of the spacing curve-outline methods.
and configuration of points rather than the shape
of the relevant anatomical or other form features. In this view, point tips, for instance, are valid
Following Bookstein's logic, therefore, care is landmarks because they possess unambiguous
required both in the placement of semi-landmark topological correspondence across specimens.
points and in analysis to control for spurious MacLeod generalized the approach even farther,
patterning, especially because outlines and other to encompass "interobject landmark corres-
complex curves are so important in stone tools. pondences" (1999:110), z:e. the use of evenly spaced
points whose placement is arbitrary with respect
Despite its own use of them, the Bookstein to anatomical traits. In Bookstein's view, such
approach to landmarks raises doubts about the points are at best semi-landmarks. Later, MacLeod
validity of semi-landmarks. Whether spaced (2008) extended the approach to eigensurface
equidistantly (either along equidistant grid lines analysis of 3D forms, calling this approach
or along radially expanding lines themselves not "eigenshape analysis" of closed curves. MacLeod's
equidistant) or otherwise by arbitrary interval, approach moots the distinction among landmark
semi-landmarks may produce spurious morpho- types and between landmarks and semi-landmarks,
metric patterning. Using hypothetical examples, and urges identical analytical treatment of
for instance, Gunz et al (2005) suggested that landmark points however defined. It also casts
semi-landmarks produce patterned deformation in doubt on the validity of the bending-energy
thin-plate spline analysis where none should occur, argument noted above.
purely as an artifact of the arbitrary location of
landmark points along edges or axes. Un- It remains for the geometrically astute to gauge
fortunately, the complex mathematical solution to the relative merits of Bookstein's and MacLeod's
this problem is not yet available in shareware views. In the interim, archaeologists have used
programs, although one apparently is under semi-landmarks (e.g., Lycett etal2oo6), a practice
development and may soon become available (Erik we follow here in part because of MacLeod's
Otarola-Castillo, personal communication, October reasoning. As above, most 3D studies to date have
Shott and Trail - Explorim: New Approaches to Utbjc Analysis: ... 201

concerned Old World Paleolithic tools, not the SCANNING TECHNOLOGY


smaller, usually more finely and elaborately made IN LITHIC ANALYSIS
bifaces ("points" in common usage) that are
abundant in the Americas but also common The ontological status of landmarks and the
elsewhere. Large, rather amorphous, bifaces such relative merits of Bookstein's and MacLeod's
as Acheulian cores or tools typically lack fine details divergent views deserve careful study. In the
of form such as well defined tips, distinct haft and meantime, it remains to consider the analytical
blade elements, basal concavities, notching, fluting, value to lithic analysis of laser scanning and the
and the like. Paleolithic tools may be characterized GM approach. As an example, we briefly describe
adequately by sets of evenly or unevenly spaced our recent work. No doubt, colleagues elsewhere
semi-landmarks (i.e., arbitrarily placed points, are undertaking studies that differ in important
sensu Bookstein). Combined differences of size, respects.
form and technology make it much more difficult
to adequately characterize points in the same way. We use a NextEnginenr portable laser scanner
Downloaded by [George Mason University] at 21:19 13 June 2016

Instead, points are best characterized using what and ScanStudio HD (high definition) software. The
Bookstein would call combinations of Type I-Ill system consists of the laser scanner itself, a data
landmarks along with semi-landmarks. Of course, cable that connects the scanner to a laptop
in MacLeod's view these distinctions are irrelevant. computer, and a revolving stand on which the scan
Accordingly, we combine Bookstein Type I or 11 subject is placed (Figure 2). Specimens are held in
landmarks (depending upon one's view of the place with rubber-tipped gripper arms or set into
Booksteinian ontological status of unambiguous modeling clay. Although our use to date is confined
locations on stone tools like the tip) with what to the laboratory, the system fits in a padded
analysts probably agree are semi-landmarks. carrying case the size of carry-on luggage and is
designed for portable use.
202 Lithic Technology, volume 35, 2

To process images and to create and manage surface setting of 8o% to 90%. Surface finish
our sometimes large data files, we use a Dell ("Finish" in ScanStudio) settings are matte or
Latitude D830 Intel Core 2 Duo laptop computer shiny. To maximize data capture, we powder each
with a T78oo processor at 2.6 GHz and 3·5 GB of specimen to diminish the effects of unwanted light
RAM. We also use a 1.81 terabyte external hard- reflection and to maximize visibility, particularly
drive for permanent storage of all scanning data, at tips and base corners and along thin edges. We
as well as high-density DVDs as a second archive do not otherwise treat specimens before scanning,
and for portability. for instance, by coating them in ammonium
chloride as Smith and Strait (2008:4-5) recom-
To produce 3D models, the system scans each mend. Most of our specimens have a matte finish,
specimen at fixed degree intervals on the rotating although fine-grained cherts and certainly obsidian
stands. Although there are set-ups that scan more would require a different setting or liberal
than one specimen at a time (e.g., Smith and Strait powdering before scanning.
2008), we are able only to scan single objects. Our
protocol was developed by trial and error, The scanning process produces a digitized
Downloaded by [George Mason University] at 21:19 13 June 2016

considering object sizes and features, the model of the specimen represented by a point cloud
reflectivity of their surfaces, and hardware/ or alternatively by a triangular mesh formed by
software capabilities. A specimen is scanned at a linking adjacent points. The density of points or
distance of 6.5 in. (16.5 cm), measured from the size of the triangles can be manipulated, altering
face of the scanner to the horizontal midline of the the resolution and overall file size. The lower the
specimen in its start position. The range for image point density or larger the triangle size, the fewer
capture is 5 in. to 9 in. (13 cm- 23 cm) which must the points and larger the triangles used to make
be taken into account when scanning, because the the mesh that represents the object in digital
specimen rotates for each interval. Our chosen format. Obviously, this yields a smaller file size.
precision setting provides accuracy to 0.005 in. for We use a high-resolution setting for triangle size
data capture. of 0.0075 in, with a range of 0.0050 in. to 0.0250
in. for capture of small, irregular details in our
Because stone tools are complex, irregular specimens.
wholes with sharp, sinuous edges and well defined
tip and base extremities, we find that a complete, The eight digitized images of a specimen from
accurate model requires successive eight-interval different angles relative to the scanner (a "scan
scans of the object, first in standing or vertical family") must be aligned to produce a complete 3D
orientation and then in prone or horizontal digitized image of the specimen. The alignment
orientation. The result is a total of 16 scans per process can be done automatically by ScanStudio
specimen. ScanStudio also takes a digital photo of via a best-fit algorithm, but we have found that
the specimen at each scanning interval, saved as a manual alignment often gives better results.
.jpg file. Alignment is done by placing a minimum of three
(maximum of nine) color-coded "pins" at
We scan at an HD (high definition) speed of 105 corresponding points on two of the eight scans
seconds for each interval; this amounts to 840 within a scan family. Corresponding points can be
seconds (14 minutes) to complete eight scans per defined by any unique patterning in color or texture
specimen. With handling, rotation and data on object surfaces. In objects with homogeneous
processing, one scan family (z:e, one specimen with surfaces, like some artifacts and certainly like
8 scan intervals) takes about 16 minutes to artifact casts that also can be scanned,
complete. corresponding points may be few; when necessary,
therefore, we use a water-soluble paint pen to mark
Two settings in ScanStudio accommodate the points on the specimen to expedite the alignment
optical properties of specimens. Surface luster process. Otherwise, marking regular intervals on
("Surface" in ScanStudio) is the laser-light the rotating platform and/or using inclusions,
exposure setting for differing texture of objects natural variation in color or texture patterning, or
represented as a percentage: o% for the darkest other unique features of the matrix of stone tools
objects and 100% for the lightest objects. For light- may suffice to identify corresponding points on two
colored specimens, our best results come from a overlapping images. In alignment, ScanStudio uses
Shott and Trail - Explorin~ New Ap_proacbes to Uthjc Analysis: ... 203

the pins to orient the separate scans and seeks averaging around 5-10 KB). The thumbnail folder
overlap between them in order to meld the separate contains two file types: .jpg Image files and
scans into one digitized 3D image. The same Shockwave Flash Object files (.swf). Every .jpg has
process is followed when aligning the vertical and an associated .swf file. For each scan interval, the
horizontal scans of a specimen. software creates a number of .jpg Image files and a
Shockwave Flash Object file (all ranging in size
After alignment, a scan family's digital models from 2-9 KB).
are trimmed to remove extraneous material (e.g.,
the platform of the revolving stand), noise, or minor The finished 3D model is saved as an .scn file.
corruptions that occurred during the scanning though it can also be saved using other file
process. The aligned and trimmed scans then are extensions. The initial scanning of a specimen at
fused, thus forming a composite 3D model, which the settings we use creates a total file size of around
removes overlapping redundant data, and results 1.5GB. However, after processing (i.e., aligning,
in a single "watertight" mesh surface. The trimming, fusing and polishing), a final model that
algorithm that drives the fusing process sometimes accurately represents the specimen reduces to
Downloaded by [George Mason University] at 21:19 13 June 2016

creates noise, particularly at very thin points of the around 250 MB on average.
object (e.g., the tip, the base corners, and the edges;
it most often appears as triangular spikes projecting
from the image). Further trimming may be CASE STUDY: APPLYING
necessary to remove this noise from the finished SCANNING TECHNOLOGY
model. In addition, there are sometimes holes in
the specimen after fusing. These are fixed by Both to test and to illustrate 3D methods, we
ScanStudio's Hole Fill feature; fortunately, such scanned and processed a set of casts of fluted
holes usually are very small and cause no bifaces kindly loaned by Mark Seeman of Kent State
macroscopic deformation of the digitized image. University. These included a Clovis specimen from
the western United States and fluted variants of
We use ScanStudio's CAD feature to establish a Clovis from other locales. The latter were
uniform orientation among all specimens, the comprised of a specimen from the Kimmswick site
artifact tip pointing upward. This is done before in Missouri, one from the Lamb cache near Buffalo,
fusing because, during the fusing process (after Gainey points from the Bostrom collection and
extraneous data have been trimmed from the from an unspecified location, a Barnes/
specimen), ScanStudio reads the extremes of the Cumberland point and a Crowfield point, both
data (i.e., the most distant points on the specimen's from unspecified locations. Figure 3 shows the
surface) and fits a "bounding box" that western Clovis and Cumberland specimens. These
encompasses the specimen. The bounding box samples have been selected to illustrate the
touches the tip and base corners and the widest part approach, not to present specific analytical
of the edges. For consistency of orientation, this is conclusions.
very helpful when we transform ScanStudio data
(in .scn file format) into .ply files for import into Coding Protocol
the program Landmark (detailed below).
Once a watertight digital model is produced, it
ScanStudio creates several types of files, which must be coded for analysis. ScanStudio
are used in the digital representation of the automatically calculates specimen volume (see also
specimen. At the start of the scanning process, the Grosman et al2oo8), a good general size measure
system creates raw folder and thumbnail folders. that is useful in reduction analysis (Shott 2005;
Raw folders contain configuration settings, DAT Shott et al 2007; Eren and Prendergast 2008).
files and Excel files. Raw-folder files provide the Landmarks and other analytical variables then
parameters for the software to graphically must be added. We save the ScanStudio .scn file of
represent the specimen; this folder runs to about each finished model as a .ply file. Then we import
700MB in size, depending on data resolution. One the .ply file into the shareware program Landmark,
scan family of eight interval scans includes 32 DAT which allows the placement oflandmark points on
files (each averaging around 20-25 KB), 32 Excel the surface of specimens.
files, and 16 configuration setting files (each
204 Lithic Jecbnologg, volume 35, 2
Downloaded by [George Mason University] at 21:19 13 June 2016

Figure 4· Our landmark coding scheme, with Clovis


specimen as example.

Figure 3. Example 3D digital models produced by


scanning: a) Clovis specimen; b) Cumber-
land specimen.
Shott and Trail - &plorin& New Awroacbes to Uthic Analysis: ... 205

In the following discussion, we make no size and shape. Like simple measurement,
distinction between the Bookstein and Macl.eod landmark placement may be subject to error. For
views, simply calling all analytical points instance, it was a concern in studies using contact
"landmarks." Our coding scheme was designed to digitizers (e.g., von Cramon-Taubadel etal2007;
capture critical points, configurations, and shapes Archer and Braun 2010:204), where the arm's
of points, reflecting both their purpose in cutting point may be placed differently by different users
and penetrating (mostly concerning tip angle and when recording landmark positions, much as
cross-section size and form) and their requirements orthogonal dimensions are subject to considerable
for secure hafting (z:e., haft area, base form). The error when measured using calipers. Error is
result (Figure 4) defined specific (called "single- possible in any coding scheme, but one advantage
point" in the program Landmark) landmarks (LM) of Landmark applied to 3D digital models is its
on each specimen. These are located at the tip (LM ability, aided by its powerful zoom function, to
So); along the edges 5 mm below the tip (LM S9, place landmarks at desired locations with great
Su); on each face equidistant between S9 and Su precision. Error studies for contact digitizers may
(LM S10 and S16; S 16 is not shown in Fig. 4); at followvon Carmon-Taubadel et al.'s design. In our
Downloaded by [George Mason University] at 21:19 13 June 2016

the point of maximum basal concavity (LM S17); study, Trail first placed landmarks following careful
base corners (LM S3, S4); distal terminus of edge study of specimens. Trail, however, is not a lithic
grinding (LM S1, S2); and flute-channel corners analyst; Shott also examined points carefully and
(LM S s-8 and S12-1s; only Ss-8 are shown in Fig. approved Trail's placement or corrected landmark
4). placement as needed.

We also defined nine landmarks at fixed Our particular combination of "single-point,"


intervals on a three-by-three grid ("mesh points" "mesh," and "curve-area" LMs and the con-
in the Landmark program). These are located in figuration of landmarks were intended to capture
the flute channel between base and farthest extent design and use variation in fluted bifaces; it is not
of edge grinding. The base (most proximal) line of suitable for all point or other tool types. LMs at
the mesh (LM M1-M3) was placed perpendicular base corners and maximum basal concavity mark
to the long axis at 1 mm above LM S17, the point of fundamental points of design and function.
maximum basal concavity. The distal-most mesh Landmarks at the distal-most extent of edge
line (LM M7-M9) was aligned with S1 and S2 grinding define the maximum extent of haft
(which often, but not always, lie at the same point lashing; those related to the flute channel, the size
along opposing edges), on the assumption that and shape of the flute, which may extend well below
hafting extended only to that point. the haft.

LM M4 -6 are mid-points within the mesh. The tip is a fixed character, but one that does
Mesh points characterize size and shape of the flute not occupy a fixed position. Instead, it is a
channel on the haft. Because we placed a set on vulnerable location susceptible to damage and
each face in the haft area, the distance between repair (e.g., Shott et al 2007). Four symmetrical
corresponding mesh points on each face estimates points on edges and blade midsections ofboth faces
thickness and cross-section size and shape. Corner 5 mm below the tip (LMs S9-11 and S16) aid in
"mesh points" may or may not coincide with LMs defining tip angle and near-tip section area and
Ss-8 or S12-15. Finally, we placed 15landmarks at form, important functional traits (Hughes
fixed, arbitrary intervals along the object's outline 1998:373-375). Reasonably but somewhat
curve ("curve points" in the Landmark program; mechanically, Shea (2005) estimated tip cross-
not shown in Fig. 4). These begin at S1, proceed section as one-half of the product of maximum
toward the base and continue along the specimen's width and maximum thickness. Crompton (2007:
circumference. Fig. 3) used a contact digitizer to capture tip and
other cross-sections of points. Our approach
The result is 14 landmarks on each face and 23 requires no geometric estimation and can capture
on the outline shared by the faces, a total of 51. All cross-section data at virtually any point on a
landmarks are placed at the same relative location specimen's long axis that the analyst desires, which
on specimens. Accordingly, relative differences contact digitizers ensure only at original data
among them describe corresponding variation in coding.
206 Lithic Jecbnplogg, volume 35, 2
Because our data are landmark-based, we have of inter-landmark distances, i.e., PCA of Procrustes
no direct measure of edge angles, curves, or cross- residuals. PCA of coordinates from full-tangent
section size or shape. However, lines linking some projection (analogous to the map projection of 3D
landmark points approximate angles and other topographic point,s onto 2D maps) projects results
geometric properties of tools. For instance, LMs into "configuration space," (O'Higgins and Jones
closest to the tip can be used to approximate tip 1998:258; 2006), z:e., the space of the original
angle, edge angles near the tip, and tip-area cross- specimens. Alternatively, PCA can be performed
section, all functionally important characters. on raw Procrustes registration coordinates.
According to O'Higgins and Jones (2006), "In
Export to Morphologika practice this choice makes little difference to the
PCA output", although it may bear upon inter-
At this point, data are suitable for morphometric pretation of centroid size, as discussed below.
analysis. Several shareware packages perform 2D
GM. We use Morphologika (O'Higgins and Jones Besides plotting the position of specimens on
2006) shareware, freely available at the SUNY cross-plots of principal components, Morphologika
Stony Brook morphometries website (http:/ I displays the configuration of landmark points at
Downloaded by [George Mason University] at 21:19 13 June 2016

life.bio.sunysb.edufmorph/), to perform 3D GM any point in the plot of principal-component axes,


analysis. whether or not empirical specimens represent
those configurations. In this way, it depicts a full,
Standard GM analysis involves three steps. Like continuous range of variation in landmark
GM itself, these steps are identified by long configurations for any pair of component scores.
expressions that practitioners reduce to acronyms. These hypothetical configurations aid
First, all objects' x,y,z landmark coordinates interpretation of components. Also, they amount
simultaneously are "registered" through to segments of theoretical morphospaces, a GM
Generalized Procrustes Analysis (GPA). This concept (discussed in more detail below) that
controls for size differences between specimens and facilitates analysis of trajectories of change in shape
focuses analysis upon shape variation. Second, that and form.
shape variation is analyzed via principal-
components analysis (PCA). Third, form variation Thin-plate splines (tps) represent specimens as
is expressed graphically in Cartesian trans- deformations of a standard or original form. In
formation grids or thin-plate splines (tps). concept, tps are infinitely thin planes whose grid
lines are aligned in perfect orthogonal orientation
GPA removes scale and orientation ("trans- on the standard specimen. Transformation of the
lation and rotation") differences, leaving shape as tps has two components (Zelditch et al 2004:134;
the sole dimension of variation. GPA produces a O'Higgins and Jones 2006). Affine components
gestalt best fit of all landmarks, rather than involve tilting or rotation of the spline but no
registration to an a priori point or baseline. Shape bending, and indicate differences between models
variation among specimens is expressed as only in orientation. Non-affine components
residuals to the coordinates of the mean shape involve deformation, such that the spline grid-lines
(Crompton 2007:3). Detailed explanation ofGPA are bent to accommodate shape change between
is best left to the geometrically more astute (e.g., models. Results depict the specimen's departure
O'Higgins and Jones 2006). However, an from the standard in both affine and non-affine
important distinction is between the two respects, i.e., 3D pattern and degree of difference
Morphologika options for projecting between the specimens.
multidimensional landmarks points onto planes for
analysis in two dimensions. One method is full- GPA of Fluted Points
tangent projection, another projection onto partial
Procrustes tangent space. We briefly consider the As above, the first step in GM analysis is
comparative merits of projection methods below. Procrustes registration via GPA. Morphologika
takes raw x,y,z coordinates (Figure sa), translates
Then PCA extracts relative warps, a somewhat and rotates them to remove the effects of
turgid and GM-specific term for the principal orientation and registers them to a common scale,
components extracted from the covariance matrix in this case using full-tangent projection (Figure
sb).
Shott and Trail- Eltplorin~ New Approaches to Lithic Analysis:,
Downloaded by [George Mason University] at 21:19 13 June 2016

Figure 5· Landmark coordinates: a, above) raw x,y,z coordinates; b, below) coordinates after rotation
to remove orientation and scale.
208 Lithic Technology, volume 35, 2
GPA uses centroid size, the square root of the Bames/Cumberland specimen fell very close to the
sum of squared Euclidean distances from each regression line; only the Bostrom specimen formed
landmark to the centroid (mean of landmark an outlier, and it is neither particularly elongated
coordinates)(Bookstein 1991:94; Zelditch et al nor nearly as great an exception as the Barnesj
2004:56). Centroid size is "a mathematically Cumberland specimen to the correlation between
natural size measure" (O'Higgins and Jones 2006), volume and centroid size.
z:e., a gestalt measure of size. More direct size
measures include weight, not available in the casts Neither the accuracy nor the precision of our
that we scanned, and volume. Fortunately, data are easily questioned. Even if they were, such
ScanStudio automatically computes object volume, constant problems could not explain inconstant
in cubic in. If centroid size is a direct size measure, results like the Kimmswick specimen's rank-order
then it should covary strongly with volume. in centroid size changing with projection method.
Much likelier, in our view, is that projection
As above, Morphologika registers data in two methods combined with how centroid size is
ways, by full-tangent projection or by partial calculated, are the source of these findings. The
Downloaded by [George Mason University] at 21:19 13 June 2016

Procrustes tangent space. Although both options conspicuous length and narrowness of the Bamesj
produce values of centroid size that pattern with Cumberland specimen, for instance, suggests that
volume, correlation is stronger using full-tangent centroid size, at least in these data, measures
projection (Table 1). In fact, parametric correlation elongation, as well as overall size.
does not reach statistical significance in
registration by Partial Procrustes tangent space, To test this idea, we computed a simple ratio of
although the small sample size advises use of maximum length to maximum width. In full-
nonparametric r., which does correlate tangent project data, this length-width ratio
significantly. correlates strongly with centroid size (r=.84,
p=.02), but not significantly with volume (r=.40,
p=.38 )(Figure 7); nonparametric r. does not alter
Table 1. Correlation (Pearson's rand Spearman's rho) the results. In Partial Procrustes tangent space
of specimen volume and centroid size. registration, neither centroid size nor volume
correlates with the ratio. Therefore, we conclude
Full tangent Partial Procrustes that, in full-tangent registration, centroid size
projection tangent space
measures elongation as much as overall size, and
that only in registration by Partial Procrustes
r= -74 p=.06 r= .60 p= ·13
r. = 7S p= .os r. = -7S p= .os tangent space does centroid size approximate
overall size. Even then, the correlation is somewhat
ambiguous (Table 1).

However, a closer look at the data prompts A simple geometric demonstration supports this
reservations about the correlation between these conclusion. Figure 8 shows two triangles of equal
variables. Cross-plots of centroid size against area, ie., of equal2D size. Dimensions in arbitrary
volume show that cases pattern differently with the units are from Dildine (1999). Triangle 1 is nearly
two registration methods (Figure 6). The isosceles in form, hence oflow elongation like most
Kimmswick point, for instance, has the smallest of our fluted specimens. (Its elongation is found
centroid size in partial Procrustes tangent space, as length divided by width or, in geometric terms,
but is larger than four other specimens using full- height h divided by the base formed by the a-c line.
tangent projection. Even more surprising is that Its elongation=o.6o.) Triangle 2 is much taller or
the Barnes/Cumberland point has the largest longer than it is wide, hence elongated ( =2.38) like
centroid size in both registrations, yet is only the the Bames/Cumberlandspecimen. Thus, triangles
third-largest specimen by volume. After obtaining differ in shape but not size. LM A bears the
this result, we reconfirmed volume measurements, coordinate value of x=o, y=o; coordinates for other
and created a new estimate of volume found as half LMs are relative to this position in units specified
the product of the multiple of maximum length, by Dildine (1999 ). Each triangle's centroid is found
width, and thickness. This estimate correlated very as the arithmetic mean of LM x and y coordinates.
strongly with measured volume (r=.91, p<.Ol). The The squared distance from each LM to the centroid
Shott and Trail - Ex.vJorin~ NeW Ap_proaches to Uthic Analysis: ... 209

300.00

•c:r;utleERLA@==-=


• BM\4SiMCKJ

• •~
~
Downloaded by [George Mason University] at 21:19 13 June 2016

tR@'ELOI •
e pAt..eVI

100.00

0.200 0.400 0.600 0.800 1.000

volume

150.00 •

125.00

"0
•~
fc 100.00 •~

0

• t:RONj@ ~

• pAN:'i1

0.400 0.600 0.800 1.000


0200
volume

Figure 6. Volume against centroid size: a) full-tangent projection; b) partial Procrustes tangent space.
210 lithic Technology, volume 35, 2

6.


5.

0 4
'fJ...
-a
-~

-3.00

Downloaded by [George Mason University] at 21:19 13 June 2016

2DO

R Sq Lnear • 0.766

1.00

100.00 150.00 200.00 250.00 300.00

Centroid

6.00


~:\..~Aft~

5.

0 4
1il...

I\IMMS@
-a
"j
c
• •~
-3.00

~
•PAINEYI ~
2. •
R Sq Linear" 0.157

1.00

0.200 0.400 0.600 0.800 1.000


volume

Figure 7· Length-width ratio in full-tangent space against: a) centroid size; b) volume.


Shott and Trajl- &plorin~ New Approacbes to Uthjc Analysis: ... 211

2.
1.
Downloaded by [George Mason University] at 21:19 13 June 2016

D 100,0

050,0

Figure 8. Centroid-size calculation in two triangles of equal area but different shape.

is found as (X2 - X1) 2 + (y2 - yy, where the centroid outliers that skew results by inflating the mean
is X1, Y1 (Zelditch et al. 2004:56). Centroid size is squared distance between landmark points.
the square root of the sum of these distances. Isolated LMs at extremes like the tip might pull the
centroid toward them, perhaps exaggerating
Centroid size of Triangle 1 is 174.9 (in whatever centroid size relative to area or volume. In Figure
units one specifies); of Triangle 2, 217.9. Thus, S's triangles, for instance, B is the most isolated
elongated Triangle 2 has a considerably larger LM. In Triangle 1, however, it is slightly closer to
centroid size despite the triangles' identical areas. the centroid than are LMs A and C, hence it
This is no surprise; centroid size "is statistically contributes slightly less than do either of those
uncorrelated with shape so long as shape changes points to centroid size. In Triangle 2, by contrast,
isometrically ( ... [z:e.] that shape does not change LM B is much more isolated and contributes much
with size)" (Zelditch et al 2004:56). In stone tools more to centroid size than any other LM, in fact
and many other things, shape might change almost as much as all others combined. Applying
considerably with size. If so, centroid size can be a the same logic to points, relatively isolated tip LMs
compound of size and other properties like shape in highly elongated specimens would yield higher
in geometric morphometric terms, as it may be in values for centroid size than their volume would
our data. suggest. Blade-area mesh points serve not only the
purpose described above, but they also reduce the
Our pilot study on Clovis-affinity specimens did probability of bias by skewing. Whatever the
not include blade-area "mesh point" LMs that we explanation, archaeologists who use GM methods
subsequently have used in coding of Hunzicker's to characterize stone tools should demonstrate, not
(2008) Folsom replicas. In Clovis data, their merely assume, that centroid size is strictly a size
absence leaves five LMs (So, S9-Su, S16) relatively measure, not a compound of several geometric
isolated at specimen tips. For biological data, Slice properties.
(2005a:19) suggested that such isolates act as
212 Lithic Technology, volume 35, 2
One reviewer suggested reasonably that the to accommodate the new positions. The resulting
ambiguous relationship between volume and thin-plate spline is a scaled graphical summary of
centroid size owes to our small sample. This all- the pattern and degree of specimens' departure
purpose concern fails to explain why our simple from the standard.
elongation measure correlates better with centroid
size than does volume. We acknowledge a possible For instance, Figure 10 compares the Clovis and
sample-size effect, as should everyone who reports Cumberland specimens in our sample (Figs. 3a,b),
analytical results from samples of less-than ideal using the former as the standard and showing
size. relative positions of comparable landmarks. By
definition, the Clovis specimen's thin-plate spline
PCA of Fluted Points is perfectly orthogonal. The corresponding spline
for the Cumberland specimen is deformed to
After GPA, relative warps are extracted by PCA, accommodate the difference in landmark
usually performed on the covariance matrix. We configurations. This view of the Cumberland
repeat that our sample is small, its patterns of specimen's spline does not clearly depict the affine
Downloaded by [George Mason University] at 21:19 13 June 2016

variation ambiguous. We present PCA results for component, but shows the deformation required
illustration, not as definite, significant conclusions. to transform the location of Clovis-specimen
landmarks to corresponding positions on the
In our data, Principal Component 1 (PC1) Cumberland specimen. (Morphologika also allows
accounted for 85.1% of variation; length and haft users to vary the number and spacing of spline grid-
size loaded positively and thickness negatively, lines, to plot splines at any intervals desired within
suggesting PC1 as a size and thinness dimension. the 3D swarm of landmark points, and to plot thin-
Principal Component 2 (PC2) accounted for 12.2% plate splines in any of the XY, XZ and YZ planes
of variation; haft-size and, to some.degree, tip- and for 3D specimens.) It depicts, for instance, the
haft-thickness measures loaded positively, and comparative thinness and elongation of the
blade-size measures negatively, suggesting that Cumberland specimen and the bending, generally
PC2 is a large-haft, thick, short-blade dimension. proximal in direction, necessary to translate Clovis-
PC2 may measure degree of resharpening. No specimen landmarks to Cumberland-specimen
other components yielded eigenvectors > 1. ones.

A cross-plot of PC1-2 shows two major Alone, thin-plate splines merely depict
dimensions of variation in this small sample deformations necessary to translate one form to
(Figure 9). From Crowfield to Kimmswick defines another. But with larger samples (e.g., that include
a range of variation about equally in the many specimens from caches or single-occupation
components. From Crowfield, the axis describes sites), types can be defined rigorously in GM
increasing size, especially length and haft size, and landmark data and the transformation from one
thickness, while simultaneously blades lengthen. to another can be summarized by the resulting
From Kimmswick to Cumberland defines a second. spline. In turn, splines can indicate which
axis of great variation on PC2 over a modest landmarks and, by extension, which parts of
positive range of PC1, indicating a progression to specimens, change in what degree and direction
longer, longer-hafted and thinner specimens. with the transformation between types. In this way,
thin-plate splines not only illustrate transfor-
Tps Analysis of Fluted Points mations but help to suggest how and why they
occurred.
Tps analysis-really, data presentation-often
begins with a standard or baseline model, usually Evaluating GM Results
what is considered a common ancestral specimen
in phylogenetic studies. For illustrative purposes, For a sample so small and opportunistic, our
choice of a standard is arbitrary; we chose the results serve merely to demonstrate GM analysis.
western Clovis specimen. As landmark points on Nor did we measure or control for different degrees
other specimens depart from the equivalent or pattern of resharpening among specimens,
position on this standard, the plane is distorted in which introduces a dimension of variation
that region from its original configuration in order independent of typology. By itself, therefore,
Shott and Trail - Explorin~ New Approaches to Uthic Analysis: ... 213

2.00000

1.00000

• PAJ£\1

~
N
0 0.00000
A.
•~
•13osmOMI
·1.
Downloaded by [George Mason University] at 21:19 13 June 2016

• Rf.'i.4SWCKI
-2.00000

·1.50000 ·1.00000 .0.50000 0.00000 0.50000 1.00000 1.50000

PC 1

Figure 9. PC1 against PC2.

analysis amounts to little more than description of Clarkson et al (2006) measured number and
differences among specimens. But larger, orientation of core facets. Lycett et al's (2006)
systematic samples can be subdivided by type, cross-beam coordinate caliper used a different
region or other criteria. After reduction allometry method to produce similar coordinate data.
is factored out (e.g., by confining analysis to haft-
area landmarks), mean landmark positions for each Contact digitizers are useful tools. Yet they
subset represent its modal characteristics. GM produce only constellations of 3D landmark
analysis, especially PCA and tps, then reveals coordinates, not whole-object models. Any
differences among the subsets, not in conventional landmark position not measured is thus not
qualitative ways,but in much more detailed, available in archived digitizer models. Except as
quantitative ways that embed geometric wire-frame approximations to complex whole
information on size, shape and form. The challenge objects, neither are they suitable for CAD
then is to develop the theory to explain those applications that can, for instance, take sections of
differences. points.

Other Options in GM Analysis


EVALUATING GM ANALYSIS
Besides the whole-object laser scanning that we
and others are undertaking, there are varieties of GM emerged in biology and related fields to
contact digitizers. These instruments use a study whole-organism form free of the severe
swinging arm to mark x,y,z coordinates of constraints of orthogonal measurement. Its
landmarks on fixed objects. They are used widely development there no doubt was painstaking and
in craniometric studies (e.g., Slice 2005b), but also characterized by progressively improved
have been used by archaeologists. For instance, approximation to whole-organism form. Similarly,
Braun et al. (2009) precisely measured flake non-orthogonal measurement in stone tools, a
platforms in three dimensions to improve precondition to the use of GM methods, developed
allometric estimates of original size of blanks. slowly over time. It can be traced back at least to
214 Lithic Technology, volume 35, 2


• I

I ~
• •
Downloaded by [George Mason University] at 21:19 13 June 2016

I•

• ~ I ••
I• ••

I
~ .
• I
I•

I


•• • • I

CLOVIS CUMBERLAND

Figure to. Thin-plate splines of Clovis and Cumberland specimens. The Clovis
specimen was used as the standard.
Shott and Trajl - &plorin~ New Awoaches to Uthic Analysis: ... 215

Holmer's (1978) pioneering study, in which a small not been contemplated, let alone addressed.
number of landmark points characterized each
specimen's partial outline form. Later, whole Applied to this problem, GM makes its own
outlines were coded (e.g., Gero and Mazzullo 1984). contribution to the essential tension between
More recently, archaeologists have developed theory and method noted above. Not intrinsically
detailed 2D landmark coding schemes (e.g., historical in nature, GM is a set of methods that
Buchanan 2006). Now 3D digital imaging and accomplishes highly detailed characterization of
landmark coding are possible. Progressively specimens while preserving the geometric
through time, then, lithic analysts have used more configuration of landmark points. This detailed
landmarks in more detailed coding schemes. characterization alone makes it possible to
distinguish between types in finer degree than ever,
Yet GM improves upon orthogonal based on geometry, not just orthogonal dimen-
measurement only by degree. Using landmark x,y,z sions. Our ability to characterize tools in such detail
coordinates, it reduces object shape to the naturally begs questions not previously raised
simplified approximation of complex, irregular about historical relationships and descent patterns
Downloaded by [George Mason University] at 21:19 13 June 2016

wholes that are points. To develop further in lithic among types. Traditionally, archaeologists treat
studies, GM analysis must be combined with CAD point types as givens, as discrete entities. Detailed
technology to record and analyze curves, angles and characterization via GM can reveal patterns of
areas. And of course, GM analysis is useless for historical continuity and descent among successive
sourcing, for use-wear analysis, and for other types that earlier studies suggested (e.g., Fischer
purposes. 1989; Bradbury and Carr 2003; Shott 2003). We
lack only the theory by which to explain these
But all analytical approaches have strengths and patterns of similarity expressed across time, and a
weaknesses. Despite its natural limitations, GM common coding scheme by which to characterize
analysis is a powerful method both to characterize successive types. Here is an example of method,
form and to analyze changes in form with GM in this case, stimulating improvements in
progressive reduction or between types. And the existing theory or new theory itself.
GM approach has other virtues that bear
consideration. Future applications of laser- The GM concept of theoretical morphospace
scanning technology and GM analysis are as easy describes all conceivable combinations of variable
to imagine as they are difficult to number. Here states or ranges of continuous variation in a class
we emphasize merely two among them, of emerging of objects. Like most GM concepts, theoretical
interest in our own research. morphospace arose in paleobiology, but applies
equally to formal stone tools like points as to
Theoretical Potential trilobites and whale skulls. In fact, one classic
exposition of it used triangular arrowpoints as its
One exciting prospect of GM analysis in example (McGhee 1999:19-20).
archaeology is the role it may play in historical
studies of point types. O'Brien and Lyman's (2003; To document the range and direction of
see also Beck and Jones 2007) cladistic study of evolutionary change, paleobiologists sometimes
late Paleoindian-age lanceolate points was followed measure how much of a theoretical morphospace
quickly by other historical studies noted above. was filled over time, and its rate and direction of
Separately and as a group, these studies are filling. Positions or variants in morphospace differ
imperfect for methodological reasons (Curnoe from neighbors in measurable degree and
2003), but also because they do not sufficiently direction. The rate at which morphospaces fill
control for allometric resharpening effects and reflects immanent tendencies to diversity, the
describe, but do not explain, historical trajectories degree of evolutionary !ability or its opposite
of change (e.g., Shott 2oo8a, 2009). Whatever (Bauhaus or other constraints), and the selective
shortcomings historical studies possess are less factors driving the evolutionary trend. In the latter
important than the valuable function they serve in sense, morphospaces are explanatory, not just
drawing archaeologists' attention to theoretical descriptive, because they focus attention on why
questions about mode and tempo of change among and how particular morphospace segments filled.
stone-tool types. Until now, such questions have Morphologika PCA's ability to generate
216 Lithic Technology, volume 35, 2
hypothetical landmark configurations helps to drift, about the complex patterns of allometric
visualize complete morphospaces, not just the variation that attend resharpening and reduction.
variants represented by empirical specimens or As much as anything, scanning technology and GM
types. MacLeod (2009) showed how placing methods provide the new data to test theory not
empirical subjects in their full morphospace contemplated before.
context aids explanation of variation in
evolutionary or other terms. They can do the same
in archaeological studies of morphospaces of stone- ACKNOWLEDGMENTS
tool types and trajectories of change in them
through time. Gilbert Tostevin, John Soderberg and Hayley
Jirasek of the University of Minnesota introduced
Documentary Potential us to scanning technology. Marcelo Cardillo of the
University of Buenos Aires, Norman MacLeod of
Although the greatest value of scanning tech- the Natural History Museum of London, Paul
nology is analytical, it has other uses as well. O'Higgins of Hull York Medical School, Erik
Downloaded by [George Mason University] at 21:19 13 June 2016

Scanning may prove an efficient way to document Otarola-Castillo of Iowa State University, and
and, in a sense, preserve the private artifact David Wiley of Stratovan Corp. offered technical
collections that never will find their way into advice on scanning or geometric morphometries.
museums (Shott 2oo8b). In the aggregate, these Ronny Islam and Michael Kim of NextEngine
collections are invaluable for the larger samples provided technical advice on the operation of
and fuller documentation of the record that they ScanStudio. Mark Seeman of Kent State University
supply. Scanning is comparatively rapid and graciously loaned us the fluted-point casts used in
accurate, as well as nondestructive, and portable this study. Three anonymous reviewers made
scanners like ours can go to the collections. several useful comments. We are responsible for
any errors or omissions.
With scanning technology and software, we can
produce accurate, full-size digital models of the
artifacts housed in private collections. The digital REFERENCES CITED
archives compiled from such efforts will have their
own intrinsic value by simple virtue of their Archer, Will, and David R. Braun
existence. But distributed on-line, they can 2010 Variability in Bifacial Technology at
dramatically increase the effective sample size for Elandsfontein, Western Cape, South
analyses stylistic, technological or typological Africa: a Geometric Morphometric
(Sumner and Riddle 2008). Long after collections Approach. Jountal of Archaeolo-
are lost, their constituent parts scattered, sold or gical Science 37:201-209.
destroyed, archaeologists can continue to excavate Beck, Charlotte, and George T. Jones
in digital archives. Depending upon the quality of 2007 Early Paleoarchaic Point Morphology
3D digital models and the availability of hardware and Chronology. In Paleoindian or
and software, it also may be possible to reverse- Paleoarchaic? Great Basin Human
engineer (recreate a physical replica in resin or Ecology at the PleistocenefHolo-
other composite material) scanned specimens. If cene Transition, edited by K. Graf and
so, then even collections and specimens that are D. Schmitt, pp. 23-41. University of Utah
lost after scanning can be, in a sense, reconstituted. Press, Salt Lake City.
Bookstein, Fred L.
1991 Morphometric Toolsfor Landmark
CONCLUSION Data. Cambridge University Press,
Cambridge, UK.
Traditional lithic analysis mostly concerns 1997 Landmark Methods for Forms without
technology, form and style. The more detailed Landmarks: Morphometries of Group
characterization of specimens available through Differences in Outline Shape. Medical
GM enables archaeologists to pose questions about Image Analysis 1:225-243.
degree and pattern of historical relationship and Bradbury, Andrew P., and Philip J. Carr
descent, about transmission modes and the role of 2003 A Method for Quantifying Size and Shape
Shott and Trail- Explorin~ New Ap_proaches to Uthic Analysis: ... 217

Dimensions of North American Hafted Clarkson, Chris, Lucio Viniciu, and Marta Miraz6n
Bifaces. In Lithic Analysis at the Lahr
Millennium, edited by N.Moloney and 2006 Quantifying Flake Scar Patterning on
M.Shott, pp. 173-188. Institute of Cores Using 3D Recording Techniques.
Archaeology Press, London. Journal ofArchaeological Science
Braun, David R., Michael J. Rogers, J.W. Harris, 33:132-142.
and S.J. Walker Crompton, Shirley
2009 Quantifying Variation in Landscape- 2007 3D Lithic Analysis.
Scale Behaviors: the Oldowan from Koobi http:// epubs.stfc ac ukfhitmeam/1752/
Fora. Unpublished ms. on file, Dept. of CMo7Crompton t.doc
Archaeology, University of Cape Town, Accessed 14 May 2009.
Cape Town, South Africa. Curnoe, D.
Buchanan, Briggs 2003 Problems with the Use of Cladistic
2006 An Analysis of Folsom Projectile Point Analysis in Palaeoanthropology. Homo
Resharpening Using Quantitative 53:225-234·
Downloaded by [George Mason University] at 21:19 13 June 2016

Comparisons of Form and Allometry. Dildine, James P.


Journal ofArchaeological Science 1999 Triangle Explorations.
33:185-199- http://mste.illinojs.edu/dildine,/beron/
Buchanan, Briggs, and Mark Collard triarea.btml.
2007 Investigating the Peopling of North Accessed 26 March 2010.
America through Cladistic Analysis of Eren, Metin 1., and Mary E. Prendergast
Early Paleoindian Projectile Points. 2008 Comparing and Synthesizing Unifacial
Journal of Anthropological Arch- Stone Tool Reduction Indices. In Lithic
aeology 26:366-393. Technology: Measures of Produc-
Buchanan, Briggs, and Marcus J. Hamilton tion, Use, and Curation, edited by
2009 A Formal Test of the Origins ofVariation W.Andrefsky, pp. 49-85. Cambridge
in North American Early Paleoindian University Press, Cambridge, UK.
Projectile Points. American Antiquity Fischer, Anders
74:279-298. 1989 Hunting with Flint-Tipped Arrows:
Burnett, Paul, and Erik Otarola-Castillo Results and Experiences from Practical
2008 Knapping on an Idea: Experimental and Experiments. In The Mesolithic in
Archaeological Projectile Point Morpho- Europe, edited by C.Bonsall, pp. 29-39.
logical Variability using Geometric John Donald, Edinburgh.
Morphometries. Paper presented at the Gero, Joan, and Jim Mazzullo
73rd Annual Meeting of the Society for 1984 Analysis of Artifact Shape Using Fourier
American Archaeology, Vancouver. Series in Closed Form. Journal ofField
Cardillo, Marcelo Archaeology 11 :315-322.
2006 Temporal Trends in the Morphometric Grosman, Leore, Oded Smikt, and Uzy Smilansky
Variation of the Lithic Projectile Points 2008 On the Application of 3-D Scanning
during the Middle Holocene of Southern Technology for the Documentation and
Andes (Puna Region): a Coevolutionary Typology of Lithic Artifacts. Journal of
Approach. Ms. on file, Depto. De Archaeological Science 35:3101-
Investigaciones Prehist6ricas y Arqueo- 3110.
l6gicas (DIPA-IMHICIHU), CONICET, Gunz, Philippe, Philippe Mitteroecker, and Fred L.
Buenos Aires. Bookstein
CastiHeira, C., M. Cardillo, J. Charlin, J.C. 2005 Semilandmarks in Three Dimensions. In
Fernicola, and J. Baeza Modern Morphometries in Phys-
2007 Analisis Morfometrico de Ios Cabezales ical Anthropology, edited by D.Slice,
Lfticos 'Cola de Pescado' del Uruguay. In pp. 73-98. Kluwer, New York.
Arqueometrfa, Segundo Congreso Henton, Gregory H,. and Stephen R. Durand
Argenti'no, pp. 1-7. Comisi6n Nacional 1991 Projectile Point Measurement and
de Energia At6mica, Buenos Aires. Classification Using Digital Image
218 Lithic Technology, volume 35, 2
Processing. Journal of Quantitative A Description, Test and Empirical
Anthropology 3:53-82. Examples of Application. Journal of
Hoffman, C. Marshal! Archaeological Science 33:847-861.
1985 Projectile Point Manintenance and MacLeod, Norman
Typology: Assessment with Factor 1999 Generalizing and Extending the
Analysis and Canonical Correlation. In Eigenshape Method of Shape Space
For Concordance in Archaeolo- Visualization and Analysis. Paleo-
gical Analysis: Bridging Data biology 25:107-138.
Structure, Quantitative Tech- 2008 Understanding Morphology in
nique, and Theory, edited by C. Carr, Systematic Contexts: 3D Specimen
pp. 566-612. Westport Publishers, Ordination and 3D Specimen
Kansas City. Recognition. In The New Taxonomy,
Holmer, Richard N. edited by Q. Wheeler, pp. 143-210. CRC
1978 A Mathematical Typology for Press, London.
Archaic Projectile Points of the 2009 PalaeoMath 101: Form and Shape
Downloaded by [George Mason University] at 21:19 13 June 2016

Eastern Great Basin. Unpublished Models.


Ph.D. dissertation, Department of http://www palass.org/
Anthropology, University of Utah. Salt modules.php?name-palaeo math&page-1,
Lake City, UT. The Palaeontological Association,
Hughes, Susan S. London. Accessed 5 December, 2009.
1998 Getting to the Point: Evolutionary McGhee, George R.
Change in Prehistoric Weaponry. 1999 Theoretical Morphology: the
Journal ofArchaeological Method Concept and Its Applications.
and Theory 5:345-408. Columbia University Press, New York.
Hunzicker, David A. Mitteroecker, Phillippe, and Fred Bookstein
2008 Folsom Projectile Technology: an 2008 The Evolutionary Role ofModularity and
Experiment in Design, Effectiveness and Integration in the Hominoid Cranium.
Efficiency. Plains Anthropologist Evolution 62:943-958.
53:291-311. Moncel, Marie-Helene, Maria Gema Chac6n, Aude
Iovita, Radu Coudenneau, and Paul Fernandes
2009 Ontogenetic Scaling and Lithic 2009 Points and Convergent Tools in the
Systematics: Method and Application. European Early Middle Palaeolithic Site
Journal ofArchaeological Science of Payre (SE, France). Journal of
36:1447-1457· Archaeological Science 36:1892-
Lin, Sam C., Matthew J. Douglass, Simon J. 1909.
Holdaway, and Bruce Floyd Morrow, Juliet, and T. Morrow
2010 The Application of 3D Laser Scanning 1999 Geographic Variation in Fluted Projectile
Technology to the Assessment of Ordinal Points: a Hemispheric Perspective.
and Mechanical Cortex Quantification in American Antiquity 64: 215-230.
Lithic Analysis. Journal ofArchaeo- Niewoehner, Wesley A.
logical Science 37:694-702. 2005 A Geometric Morphometric Analysis of
Lycett, Stephen J. Late Pleistocene Human Metacarpal 1
2009 Quantifying Transitions:. Morphometric Base Shape. In Modern Morpho-
Approaches to Palaeolithic Variability metries in Physical Anthropology,
and Technological Change. In Source- edited by D.Slice, pp. 285-298. Kluwer,
book of Palaeolithic Transitions, New York.
edited by M. Camps and P. Chauhan, pp. O'Brien, Michael J ., and R. Lee Lyman
79-92. Springer Science, London. 2003 Cladistics and Archaeology.
Lycett, Stephen J., N. von Cramon-Taubadel, and University of Utah Press, Salt Lake City.
R.Foley O'Higgins, Paul, and Nicholas Jones
2006 A Crossbeam Coordinate Caliper for the 1998 Facial Growth in Cerocebus torquatus:
Morphometric Analysis of Lithic Nuclei: an Application of Three Dimensional
Shott and Trail - E!vJorine New Approaches to Uthlc Analysis: ... 219

Morphometric Techniques to the Study 2009 Theory in Archaeology: Morphometric


of Morphological Variation. Journal of Approaches to the Study of Fluted Points.
Anatomy 193:251-272. Paper presented at the 74th meeting of the
2006 Tools for Statistical Shape Analysis. Hull Society for American Archaeology.
York Medical School, Atlanta.
http:/ fhyms.fme.googlepages.comf Shott, Michael J., David A. Hunzicker, and Bob
resources. Patten
Accessed 13 October 2009. 2007 Pattern and Allometric Measurement of
Renfrew, C. Reduction in Experimental Folsom
1973 Before Civz1ization: the Radiocarbon Bifaces. Lithic Technology 32(2):203-
Revolution and Prehistoric Europe. 217.
Knopf, New York. Slice, Dennis E.,
Riel-Salvatore, Julien, Myungsoo Bae, Peter 2oosa Modern Morphometries. In Modern
McCartney, and Anshuman Razdan Morphometries in Physical Anth-
. 2003 Palaeolithic Archaeology and 3D ropology, edited by D.Slice, pp. 1-45.
Downloaded by [George Mason University] at 21:19 13 June 2016

Visualization Technology: Recent Kluwer, New York.


Developments. Antiquity 76:929-930. 2005b (editor) Modern Morphometries in
Shea, John J. Physical Anthropology. Kluwer,
2005 The Origins of Lithic Projectile Point New York.
Technology: Evidence from Africa, the Smith, Nicholas E., and Suzanne G. Strait
Levant, and Europe. Journal of 2008 PaleoView3D: From Specimen to Online
Archaeological Science 33:823-846. Digital Model. Palaeontologia
Shott, Michael J. Electronica 11(2),
2003 Time as Sequence, Type as Ideal: Whole- http:ffpalaeo-electronica.org/2008 2/
Object Measurement of Biface Size and 134/indexjhtmJ.
Form in Midwestern North America. In Accessed 17 December 2009.
Multiple Approaches to the ·Study Sumner, T. Alexandra, and Andrew T. Riddle
ofBifacial Technology, edited by M. 2008 A Virtual Paleolithic: Assays in Photo-
Soressi and H.Dibble, pp. 251-271. grammetric Three-Dimensional Artifact
University of Pennsylvania Museum Modelling. PaleoAnthropology 2008:
Press, Philadelphia. 158-169.
2005 The Reduction Thesis and Its Thulman, David .K.
Discontents: Review of Australian 2006 A Reconstruction of Paleoindian
Approaches. In Lithics "Down Social Organization in North
Under:" Australian Perspectives Central Florida. Unpublished Ph.D.
on Lithic Reduction, Use and dissertation, Dept. of Anthropology,
Classification, edited by C.Clarkson Florida State University, Tallahassee.
and L.Lamb, pp. 109-125. British Tompkins, Charles N.
Archaeological Reports International 1993 Classifying Clovis Points: a Study
Monograph Series 1408, Archaeopress, in Metric Variability. Unpublished
Oxford. MA Thesis, Department of Anthropology,
2008a Darwinian Evolutionary Theory and University of Arizona, Tucson.
Lithic Analysis. In Cultural Trans- von Cramon-Taubadel, Noreen, Brenda C. Frazier,
mission and Archaeology: Issues and Marta Miraz6n Lahr
and Case Studies, edited by 2007 The Problem of Assessing Landmark
M.O'Brien, pp. 146-157. Society for Error in Geometric Morphometries:
American Archaeology Press, Wash- Theory, Methods, and Modifications.
ington, D.C. American Journal of Physical
2008b equal o nil rootht w ded I e vsbr cted: A Anthropology 134:24-35.
Proposal for Conservation of Private Wynn, Thomas and F. Tierson
Collections in American Archaeology. 1990 Regional Comparison of the Shapes of
SAA Archaeological Record 8(2): Later Acheulian Handaxes. American
30-35· Anthropologist 92:73-84.
220 lithic Technology, volume 35, 2
Zelditch, Miriam L., Donald L. Swiderski, H. David
Sheets and William L. Fink
2004 Geometric Morphometries/or Bio-
logists: a Primer. Elsevier, San Diego.
Downloaded by [George Mason University] at 21:19 13 June 2016

You might also like