DIGITAL CRABTREE:
COMPUTER SIMULATION OF FOLSOM FLUTING

Tony Baker 9/1/99

ABSTRACT

This paper is a status report of computer simulation work of the Folsom fluting process. The simulation method was Finite Element Analysis. The simulation work indicates there are two modes from which channel flakes can be made. The energy-rich mode requires large angles of blow (AOB) to make channel flakes and the energy-poor mode requires small AOBs. Acceptable channel flakes cannot be made with AOBs falling between these two modes. The mechanics of crack propagation between the two modes are different, and therefore so are the terminations of the channel flakes. The author suggests that the energy-rich mode is the one employed by the Folsom people. Both energy modes can produce overshot flakes and the location of the anvil is a critical fact or in reducing or eliminating the overshot flake. Finally, the simulation offers a functional reason for leaving the second face unprepared while fluting the first face.

INTRODUCTION

The Folsom Point is possibly one of the finest examples of flaked stone art in the world. Since 1926 when it was first reported, it has been collected, studied, and replicated. In 1999, its manufacture is being computer simulated.

The genesis of this paper was the 1997 Folsom Workshop held in Austin, Texas. This Workshop brought together many knappers who could replicate the Folsom point. Surprisingly, each had a different technique of fluting and all experienced high failure rates during fluting. Additionally, it was obvious to all the attendees that the modern knapper had not discovered and/or mastered the fluting technique used by the Folsom people. The workshop raised the following question: "Is there a simple, successful way to flute that all the Folsom people used or is it so special a skill the only a few Folsom craftsman were making all the points?"

In the fall of 1997, this question caused the author to conceive of the idea of modeling the fluting of a Folsom point with a computer. Finite Element Analysis software was acquired in February and the testing of the software and the concept began the same month. The software was tested by replicating the flake-creation research of Pelcin (1996). The testing was successful and that phase of the research was documented as a "web" page to allow review by both the archaeology and engineering communities.1 In February of 1999, the software was focused on the subject of channel flake removal and this is a status report of that effort as of September 1999.

COMPUTER MODELING

The mechanics of flake creation is an extremely difficult subject. The reasons for this are numerous, but one reason looms larger than the others. The removal of any given flake in the real world cannot be repeated because flake removal is a destructive process. The knapper cannot recreate the exact event. Even if the knapper could recreate exactly the same core morphology and angle of blow, the lithic material would still be different.

Some of the difficulty of not being able to repeat an experiment can be overcome with experimental archaeology by attempting to replicate exact morphologies, lithic materials and force applications. However, the fact remains that even in the laboratory these experiments cannot be repeated exactly. Additionally, experimental archaeology is expensive, time consuming and lacks glamour. Therefore, to date there has been little of this type of work performed (Bonnichsen 1977, Faulkner 1972, Pelcin 1996, and Speth 1972).

Computer modeling is a form of experimental archaeology that can overcome some of the problems of real world experiments. It does not suffer from inexact repetition, nor is it expensive. It can be time consuming and probably also suffers from the lack of glamour. Computer modeling can repeat an experiment exactly, as many times as desired, and it yields exactly the same result. For these reasons, computer modeling is one of the most powerful tools in the manufacturing industry and hard sciences.

To model the mechanics of channel flake removal the author used a computer program of the Finite Element Analysis (FEA) category. FEA is a mathematical method of simulating how an object will perform in service. A virtual object (model) is constructed in the computer with simulated loads and boundary conditions. The model is divided into many small finite cells and each cell is connected to the surrounding cells. If a cells moves, it causes the adjacent cells to also move. FEA software calculates the response of all these cells and thus the object to the loads and boundary conditions. It solves thousands of complex calculations, requiring several lifetimes if done by hand, and provides the user with a deep insight to the behavior of the object. The actual software used was the Linear Stress Model, offered by the ALGOR Corporation of Pittsburgh, PA.

History Matching

Computer modeling begins with creating a model and then testing the model to see if it can replicate (history match) past, real events. If a model cannot replicate real events, one cannot trust it to predict future events.

The history selected to be matched was the experimental flake creation research performed by Pelcin (1996). Pelcin's research involved creating hundreds of glass flakes from different cores with different morphologies, angles of blow and other conditions. During his research he discovered that he was creating two different types of flakes under very similar conditions. This was most intriguing to the author and so this was the portion of Pelcin's research that was chosen for the history matching. If the FEA model could match these events, it could be trusted to predict the mechanics of channel flake removal.

A very detailed account of the history matching of Pelcin's research can be found on the "web" and, therefore, will not presented here.2 However, a few salient points from the history matching are listed below because they were used in the predictive stage of the computer modeling.

  1. The history matching was performed with a two-dimensional FEA model. Pelcin (1996:329) investigated several core shapes but most of his work was with edge cores. These edge cores were made of plate glass with dimensions in the range of 3 by 5 inches by ˝ inch thick. For the history matching a 2-D model of approximately 4000 cells, ˝ inch thick was created.
  2. Physical properties of window glass were used during the history matching and also in the predictive phase.3
  3. The modeling was qualitative and not quantitative. A 1 pound force was always used and the crack was always forced to propagate in the direction of maximum tensile stress. There was no concern if the maximum tensile stress of glass had been reached or not. Increasing the force to 2 pounds increases the stress within the model, but it does not alter the direction of maximum tensile stress.
  4. Figure 1. Terminology - Theoretical Platform Thickness (TPT), External Platform Angle (EPA), and Angle of Blow (AOB).
    Pertinent terminology used throughout this paper is depicted in Figure 1. The angle of blow (AOB) is measured from a plane parallel to the dorsal face of the flake to be removed. This is the same convention used by Cotterell and Kamminga (1987:676). The theoretical platform thickness (TPT) is measured from the dorsal face. This is Pelcin's convention (1996:110). External platform angle (EPA) is also Pelcin's convention (1996:101). The EPA of the preform in the predictive phase has not been varied to date. It has always been 55 degrees.

Cotterell and Kamminga (1987:685) divided flake formation into three phases; initiation, propagation, and termination. Propagation accounts for 90% or more of the travel of most cracks. To simplify the model, the propagation phase was chosen to be modeled. The initiation phases (hertzian, bending, or wedging) or the many termination phases were not considered in the history matching. Even with this simplification the history matching of Pelcin's work was extremely successful. Flake mass, length and the two different flake types were replicated. The two flake types turned out to be the results of two propagation modes, "stiffness-controlled" and "compression-controlled", described again by Cotterell and Kamminga (1987:692). Within this paper these two modes will be termed "energy rich" and "energy poor", respectively. The reasons for this change in terminology will become evident later in the paper.

Predictive Modeling

Predictive modeling is only as good as the model. Any model can predict future results. For a model to be trusted to predict accurate results, it must be able to history match past events. In addition, the model can be relied on more if the predictions are of similar events and under similar conditions as those in the history match. The predictive phase of Folsom point fluting is very similar to the history-matching phase of Pelcin's work. As a result, the insight gained in this computer modeling can be considered extremely reliable.

There is only a single computer model of a Folsom preform's morphology discussed in this paper. Its derivation can be visualized by Figure 2. It represents the section of the preform from which the channel flakes are removed. It is 50.8 mm long, 12.7 mm wide and 6.4 mm thick. Since the computer model is a 2-D model the two dimensions modeled are length and thickness. In Figure 3, these two dimensions are divided into 5000 cells (200 by 25 cells). The proximal end (striking platform) is on top with a 55 degree EPA. The distal end is beveled to 70 degrees.

Figure 2. Development of the model (preform) used in the computer. The model (right) represents a slice (middle) cut from a Folsom preform (left).</TD>Figure 3. Edge view of the computer model (preform). In the left image the model consists of 5000 cells, each 12.7mm wide (deep into the page). In the right image the distal end is standing on an anvil that permits only rotation. The upper support (upper right side of preform) allows all but horizontal movements.

A real preform will not stand up by itself while being fluted and neither will a virtual one in the computer. In the computer the distal end is supported on an anvil permits it to rotate but not to move horizontally or vertically. A second, upper support is mounted on the face that is not being fluted and it restricts only horizontal movement. Rotation and vertical movements are permitted at the upper support. Refer to Figure 3.

During the history matching of Pelcin's work, it was discovered that the force had to be coupled with elasticity to history match one of Pelcin's two flake types. This is the elasticity of the indentor when it contacts the preform. When a steel ball bearing is dropped on a steel plate, it bounces. It does this because steel is elastic just like rubber. Bone, antler and rock indentors are also elastic.

Figure 4. Edge view of the computer model (preform) with a crack. The 1 pound force and two springs represent the indentor. The springs replicate the elasticity of the indentor.
This elasticity was modeled with two springs as depicted in Figure 4. The 1 pound force in association with the two identical springs represents the elastic indentor in the model. During impact, the preform and the indentor compress or pull away from each other. This compression effectively reduces the force on the preform for a short time, and then both return to their original shapes. If the preform should initiate a crack and propagate it during the time the force is reduced by this compression, then a totally different shape of flake is created compared to the flake that is created under full load.

Two springs are necessary because the 1 pound load has a vertical and a horizontal component. In the computer these springs are anchored to the earth. If the force was only vertical then only a vertical spring would be necessary. The vast majority of the modeling discussed in this paper was done with the springs set to 150,000 pounds per inch. It will be noted, when they were set to a different value.


TWO FLAKE PROPAGATION MODES

Pelcin identified two different flake types in his research. The history matching of his work with the FEA model verified there were two different flake types, resulting from different propagation modes. According to Cotterell and Kamminga (1987:691), "cracks in any solid can grow either in a stable fashion under mechanical work done by an external force or unstably under the release of elastic strain energy (stored potential energy)." Reiterating this in their words, in the stable mode "cracks … grow only as the load is increased." In the unstable mode, "cracks … quickly will accelerate to high velocities of propagation" (1987:692) which mean they outrun the indentor.4 In this paper, the stable mode is the energy-poor mode because the crack keeps requiring more energy to propagate. The unstable mode is the energy-rich mode because the energy to propagate the crack is present (deformation of the core) before the crack begins. Once its starts it requires no additional energy.

Modern flint knappers create flakes in both modes. The energy-poor mode of propagation occurs when the knapper uses small AOBs to remove flakes. The energy-rich mode occurs at larger AOBs. Crabtree (1968) manufactured Mesoamerican blades in the energy-poor mode. Hard hammer, biface reduction is generally done in the energy-rich mode.

To elaborate on the two modes, consider Figures 5 & 6. Figure 5 is a contour map made from numerous simulation runs using the Folsom preform in Figure 3 and a 1 pound force with elasticity. It is a plot of potential energy as TPT and AOB are varied. On the contour map a maximum ridge of energy runs from lower left to upper right. On both sides of the ridge the energy values fall off. Figure 6 is a profile drawn from Figure 5 at a constant AOB of 30 degrees. Potential energy is plotted against TPT. TPT can be considered a proxy for preform stiffness. The preform is stiffer at large TPTs as compared to small TPTs. Also, the preform becomes less stiff as a crack forms and begins to grow. Therefore, as a crack grows in the preform, it is equivalent to moving to the left (smaller TPT) in Figure 6. If the crack starts at 2.0 millimeters, moving to the left requires more energy to propagate it. This crack is energy poor and requires more energy to continue. The flake must be continually pushed to finish it. Compare this to a crack that begins at 1 millimeter. At 1 millimeter, moving to the left requires less energy than is present in the core so the crack propagates itself. It is energy rich.

Figure 5. Contour map of the potential energy added to the computer model (preform) for various TPTs and AOBs. This is the energy prior to the initiation of the crack. Note the ridge of maximum energy running from lower left to upper right.

Figure 6. Potential energy added to the computer model (preform) for various TPTs. AOB is constant at 30 degrees. Cracks that initiate on the left side of the peak energy (1.6mm) are energy rich and will propagate without additional force. Cracks beginning on the right side of the peak energy are energy poor and will continue to require energy to propagate throughout the creation of the flake.

Crack Propagation Speed

Crack propagation speed is believed to be different between the energy-poor mode and the energy-rich mode. The speed of the crack in the energy-poor mode cannot be faster than that of the indentor. The flake must be pushed off by the indentor. The history matching of Pelcin's research indicates the flake moves away from the indentor in the energy-rich mode. Intuitively, this means the crack moves faster than the indentor and therefore must travel faster than a crack in the energy-poor mode.

If one assumes that flakes can be made in either energy mode by pressure, indirect percussion or percussion, then there is data to support the two modes having different velocities.

Pressure flaking when first considered appears to have no velocity associated with the indentor. Yet, flakes are made by this method. One of the most extreme examples of pressure flaking is Crabtree's Mesoamerican blade manufacture. These blades were made with a chest crutch and the force was applied almost vertical, or with an AOB near zero. There is no doubt that these were being made in the energy-poor mode. Crabtree reported: "When I am in good form and am familiar with the material, I am able to stop the blades at will and, occasionally, even to leave them still adhering to the core." (1968:472). If he was able to stop the propagation of the crack, he must have been pushing the blade from the core.

Crabtree's blade manufacture was timed with high speed photography and elapsed times measured at 1/1,250 of a second (1968:474). If one assumes his blades were near 10 centimeters long, then this is a speed of 125 meters per second (m/s). Since Crabtree was applying pressure with a chest crutch, where did this velocity come from? The velocity was that of the chest crutch expanding as the crack was released. Prior to the blade's release, the crutch had been compressed by Crabtree's weight as he applied load to the core. Therefore, the velocity associated with energy-poor flake propagation is the velocity of the indentor plus the velocity of the indentor decompressing. In pressure flaking the velocity of the indentor is zero.

Cotterell and Kamminga (1987:680) reported average pressure flaking crack velocities of 124 m/s. The energy mode could not be determined from the paper, however, it would be difficult to accept that these were created in the energy-rich mode because of the previous Crabtree data. Hutchings (1999) measured "fracture wing" velocities in experimentally created obsidian pressure flakes. He reported velocity values ranging from 152 to 615 m/s with an average of 371 m/s (1997:50). The energy mode was not known, but personal conversations with Hutchings indicated that flakes were probably made in both modes.

From the above data, I suggest that the velocities in the range of 100 to 150 m/s are associated with the energy-poor mode. Flake propagation velocities much higher than 150 m/s probably represent the energy-rich mode. To put this in perspective, 100 m/s is 224 miles per hour and sound travels at approximately 335 m/s in air.

There are numerous measurements of velocities much greater that 150 m/s. Cotterell and Kamminga reported velocities 600 to 1000 m/s resulting from dropping steel balls on glass (1987:680). Hutchings reported average velocities ranging from 667 to 1037 m/s for his experimental obsidian flakes manufactured under indirect and direct percussion (1997:50). In both of these cases the energy mode is not known, but it is strongly suspected that these represent the energy-rich mode.

Two Energy Regions

Predictive modeling with the FEA software indicates that channel flakes can be removed in both energy-rich and energy-poor modes. Figure 7 indicates the location of these two modes. The region immediately above the zero degree AOB produces the energy-poor channel flakes. Notice the narrowness of the range of AOBs. In this mode, the knapper has very little room for error. Possibly this region could be wider with different values for the variables in the computer model. However, to date the author has only investigated the model described in the section on predictive modeling.

Figure 7. Contour map indicating the location of the energy rich and energy poor regions. In between the rich and poor energy regions is an area where flakes will not run for any appreciable length and channel flakes cannot be made.

Immediately above the maximum energy line, lies the region of the energy-rich channel flakes. This region permits a larger range of AOBs, but the TPT is critical in this region. For the conditions investigated to date, a TPT much larger that 1 millimeter results in overshots. Again, this transition from normal to overshot flakes may move with different variable values in the computer model.

The region immediately below the energy-rich region does not produce acceptable channel flakes. They do not run and have the appearance of the flakes in Figures 8k and 8l. In essence, there is a "no man's land" between the two regions where flakes do not run and acceptable channel flakes cannot be made. To reiterate this point, there is not a continuum between the two energy modes that will produce acceptable channel flakes.

Figure 8. Various synthetic channel flakes (shaded) created with the computer model. Top row was created in the energy rich mode and the bottom row was made in the energy poor mode. a) K=10E4 #/inch (elasticity of indentor), AOB=40 degrees, TPT=1.86 millimeters. b) K=15E4, AOB=40, TPT=0.93. c) K=10E4, AOB=60, TPT=3.72. d) K=15E4, AOB=40, TPT=2.48. e) K=15E4, AOB=50, TPT=1.24. f) K=15E4, AOB=60, TPT=1.24. g) K=0, AOB=1.25, TPT=0.93. h) K=0, AOB=2.5, TPT=0.93. i) K=0, AOB=10 TPT=3.10. j) K=0, AOB=2.5, TPT=1.86. k) K=0, AOB=20, TPT=0.93. l) K=0, AOB=15, TPT=1.86.

Overshot flakes that dive into the body and exit out the back of the preform can also be made in both regions. This is extremely important as a test of the predictive model because overshot flakes are very common in the real world.

As stated previously, all the information in this section was developed with a unique set of variables, all of which were detailed in the Predictive Modeling section. Some of these are core morphology, support locations, lithic material properties, elasticity values (indentor and anvil hardness). Changing most of these variables will shift the locations of the two energy regions. Some variables will effect the shift more than others will. However, at this writing it is believed that there will not be significant changes in the pattern presented in this paper.

One variable that does not effect the shape of the channel flake or the location of these fluting regions is the force of 1 pound. If all the other variables are held constant, increasing the 1 pound force will not alter the location of the maximum energy line or the shape of the flakes. It will raise all the potential energy values in Figure 5, but it will not change the shape of the contour map. The predictive model employed in this research indicates that the magnitude of the force does not effect the shape of the channel flake. This statement assumes that there is enough force applied to initiate a crack in the energy-rich mode or push it in the energy-poor mode. Reiterating, direction of the force (AOB) and elasticity of the indentor does effect the shape of the flake.

Flake Termination Signature

Figure 9. Distal fragment of a Folsom preform with a flute scar that terminates in a deep step. A deep step termination is a characteristic of an energy rich crack propagation. (Baker collection).
Crack propagation speed is one of the differences between the two energy modes. Another difference is often the termination of the flake. Based on the predictive modeling most channel flakes beginning in the energy-rich mode exhaust that energy, and then the energy-poor mode finishes the crack's propagation. In different words, during the energy-rich mode the crack outruns the indentor until it stops in the body of the preform. Next, the indentor catches up to the flake and continues to push it to completion (energy-poor mode). When this happens there is usually a sudden direction change of the travel of the crack.5 Classic examples of this direction change are step and hinge terminations. Figure 9 depicts a step termination on a distal fragment of a Folsom preform. Figure 8b is an example of a step termination created with the computer model.

Deep hinge or step terminations cannot be produced when a channel flake is being created in the energy-poor mode. The exception to this is when there is a flaw in the material. The energy-poor crack trajectory contains only gentle changes in direction and is free of abrupt changes. Energy-poor trajectories move toward the surface as if they are feathering out. There usually is a thin step at the end of this feathering, but this is a because the channel flake is so thin it encounters a sudden change in cross-section created by flake scars on the surface. Figures 8g & 8h are examples of the channel flakes propagating in the energy-poor mode until they near the edge and break out. Notice that the trajectories are slowly moving toward the surface. The final step termination is an artifact of the computer model that is made up of cells that are about 0.25 millimeters wide. When the channel flake approaches 1 millimeter thick, a crack across one cell is a 25% reduction in the flake's cross-section. This sudden change in cross-section causes a sudden increase in the tensile stress and the crack goes to the edge. This phenomenon is similar to encountering a surface flake scar causing a sudden reduction in cross-section. Admittedly, the visual differences between the step termination of an energy-rich channel flake and an energy-poor one are subtle but they still are created by different mechanisms. Deep step or hinge terminations should be the result of the energy-rich mode. Thin step terminations are difficult to assign to a specific mode.

The energy-rich mode can also produce channel flakes that feather out. Figure 8c is an example. According to the computer model, these occur only at large TPTs and extremely large AOBs.

Figure 10. Distal fragment of a finished Folsom point. The arrow points to the origin of a pressure flake that removes the deep step termination of the channel flake. This removing of the step is common on most finished Folsom points. (Baker collection).
Many finished Folsom points exhibit what appears to be a feathering out of the flute scar. However, on closer inspection, these flute scars do not feather out. They actually terminate in a step and the Folsom knapper has removed the step with a small pressure flake coming in from the lateral edge. Figure 10 is a distal fragment of a finished Folsom point illustrating this step removal. Step terminations that have been removed as in Figure 10 are very common on distal tips of finished Folsom points. Occasionally, their removal can be found on the body of unfinished preforms, but these are rare since most preforms were abandoned after an overshot and not a step termination. Feather terminations are also rare.

Since step terminations are common channel flake terminations, it can be argued that the Folsom knappers were fluting their preforms in the energy-rich mode. This is definitely true in some cases, yet it is possible that some Folsom knappers were using the energy-poor mode. There are occasional feather terminations in the archaeological record. Assuming that both energy modes were employed by the Folsom people to flute their preforms, then the energy mode might be a band (cultural) identifier. Most likely, the same energy mode would be used within a single group since presumably they learned to flute from each other. So if the modes could be identified in the archaeological record, i.e. flute scars and channel flake characteristics, then some insight into band mobility might be gained.

At the Folsom Workshop there were definitely two groups of knappers as defined by the energy modes. The 1997 Workshop was dominated by the energy-poor mode knappers. They may have arrived at this mode independently or they may have been influence by the writings of Crabtree (1966), Flenniken (1978 ) and Sollberger (1985) who all fluted in the energy-poor mode. Possibly, Crabtree influenced Flenniken and Sollberger.

THE OVERSHOT FLAKE

Figure 11. Folsom preform replica destroyed by an overshot flake. The left image shows the channel flake that traveled about half the length of the preform before it cut the preform into two pieces. The top of the preform has been flipped over in the right image to show the channel flake scar. (Made during the 1999 Folsom workshop. Knapper is unknown.)
If the Folsom preform's platform is properly struck, fluting is similar to the old saying about passing the football, every time a pass is attempted one of three things can happen and two of them are bad. The good outcome from fluting is the removal of a long channel flake so the manufacture of the point can proceed. Examples are Figures 8a, 8f and 8g. The second and undesirable outcome is that too short a flute is removed and the preform has to undergo significant modification before making another fluting attempt. See Figures 8b, 8c, and 8h. The third, and dreaded, outcome is the overshot flake (Figure 11) which dives into the body of the preform and destroys it by breaking it in half. Simulated overshots created with the model are shown in Figures 8d, 8e, and 8i.

Assemblages from Folsom campsites usually contain performs that were destroyed by an overshot flake. At Lindenmeier overshots and preforms splitting6 were the most common failures (Wilmsen 1978:104). At the Folsom Workshops overshots were the most common fatal failure. It can occur during the fluting of either face, but it is most frequent during the fluting of the second face when the preform is the thinnest. Occasionally, the overshot occurs after the channel flake has run an appreciable distance and the knapper is able to salvage a short Folsom point. Sometimes the overshot occurs so early, that the distal end can be reworked for another fluting attempt. Most of the time, the preform is not salvageable.

The overshot flake is an interesting phenomenon. Is it the natural result of certain conditions that are repeatable, such as AOB or TPT? Or, is it the result of uncontrollable factors such as a localized weakness in the material, multiple contacts with the indentor or possibly the slipping of the preform on the anvil? I believe most knappers would agree that it is a predictable and repeatable process. They know that if the TPT is too thick an overshot will occur. Additionally, individuals who replicate Clovis material, can purposely create the overshot flake (outré passe) at will. This is done in the energy-poor mode with very low AOBs.

Mechanics of the Overshot Flake

Imagine a 2 inch thick phone book with the pages glued together. Now, imagine the phone book is bent so the front cover is concave in shape and the back cover is convex. The pages in the back of the book will be in tension and the ones in the front will be in compression. The basic premise of this FEA modeling is that a crack will propagate toward the maximum tensile stress. If the phone book were cracking down the middle, the crack would turn toward the back of the book where the pages are in tension.

Figure 12 consists of four diagrammatic sketches of the flexing of the preform in Figure 8e as it is being fluted. The actual horizontal displacements during the fluting process are on the order of 0.00003 millimeters. To visually present this flexing of the perform, these horizontal displacements have been exaggerated 350,000 times that of the vertical in the Figure. The dotted line represents the channel flake as it is removed. The preform's distal end (bottom) rests on an anvil, but it is free to rotate as if attached to a piano hinge. The upper support is located at the point the bent preform crosses the at-rest position (vertical line) in Figure 12a. The upper support only prevents horizontal movement. The AOB is 50 degrees, so the force is downward and to the left on the channel flake.

Figure 12. Diagrammatic sketch of a Folsom preform (Figure 8e) flexing while being fluted. The horizontal axis has been exaggerated 350,000 times. The dotted line is the channel flake and the dark solid line is the preform. The thin solid line is the at-rest or neutral position. In a) the crack is 5.1 millimeters into the preform; b) 15.2 mm, c) 30.5 mm, and d) 32.8 mm. At d) the crack begins the overshot dive to the back edge.

The purpose of the images in Figure 12 is to demonstrate the motion of the preform. In 13a the crack travel has not passed the upper support and the main portion of the preform is bent to the right of the at-rest position. The fluting face and associated left half of the cross-section of the preform is in compression. The right half is in tension. The compression in the left half prevents the crack from traveling into the body of the preform.

As the channel flake travels past the upper support, Figure 12b, the preform begins straightening out and then starts bending to the left, Figure 12c. Bending to the left places the left half of the cross-section in tension. Finally, in Figure 12d, the tension becomes so great that the crack changes direction and begins moving into the body of the preform. This is the start of the overshot. Once the crack starts in this direction the cross-section becomes thinner and there is no stopping the crack.7

The trajectory of the crack in the preform in Figure 8d has a special character. Midway down the preform, the crack turns into the body, crosses the center of the cross-section, and then turns down again and runs a significant length before overshooting. The shape of this channel flake is not seen in the archaeological record, nor has it been produced at the Folsom Workshops. In the real world, the crack never would turn down the second time. It would have continued straight to the back edge and cut the preform in two. The reason this computer simulated channel flake has this morphology is the rectangular cross-section of the preform, Figure 2. Real-world preforms tend to be lenticular and therefore are not as stiff as a rectangular cross-section. This additional stiffness reduces the flexing of the preform depicted in Figure 8d and retards or eliminates the overshot flake. The simulated channel flake in Figure 8d is correct for the cross-section of the preform.

Folsom preforms usually have minimal modification on the second face prior to fluting the first face. Sometimes the second face was not modified at all and was just the ventral side of a large flake. Flenniken (1978:479) has suggested that doing this "… minimized time and energy expended in the production of a Folsom" point. Patten (1999:96) rejects the concept of "… gambling on failure…" and suggests that it adds strength to the preform while fluting the first side. Removing any mass from the second side makes the preform more flexible and more inclined to overshoot. The computer simulated flake in Figure 8d supports Patten. The Folsom people were apparently increasing their chances of success during the fluting of the first face by not removing material from the second face.

Beveled Distal End of the Preform

Figure 13. Distal end of the computer model of a Folsom preform. The black triangle is the anvil. In a), the anvil is on the face to be fluted. In b), the anvil is moved in from the face to be fluted. During computer simulations, overshot flakes could only be made with the configuration in b). With the anvil on the face to be fluted, overshot flakes could not be created.
If one assumes the overshot flake is a natural result of certain conditions, then the FEA model must be able to simulate it. For many months, the author tried to make overshot flakes with the FEA model with no success. It was not until the anvil was moved away from the edge of the face to be fluted that overshots became possible. Overshots were not possible to model with the anvil located as in Figure 13a. With the anvil positioned at 1.52 millimeters from the edge being fluted, as depicted in Figure 13b, overshots became possible. Figures 8d-f (energy-rich) and Figures 8i-j (energy-poor) depict some of the overshots that were created with the anvil moved as in Figure 13b.

In the FEA model, the anvil can be repositioned without altering the morphology of the preform or changing any other parameter. Obviously, this would not be possible in the real world. The contact point of the preform on the anvil is the most distal part of the preform. To purposely position this contact point in relation to the fluting face, the knapper must bevel the distal end. The other option is to utilize an existing contact point, and elect to flute a certain face. The distal end would then have to be beveled to flute the second face.

The purposeful beveling of the distal end has been recognized in the past. Flenniken (1978:475) reported observing it in the Lindenmeier assemblage. Crabtree (1966:13) thought the bevel was "…prepared on the side opposite that to be fluted" and it purpose was to permit "… the fluting to be completed without the channel flake contacting the anvil or support." Based on the FEA research supporting this paper, the bevel bias should not be on the side opposite the one being fluted, but on the one being fluted. This should reduce the overshot failures. Gene Titmus' (1999: pers. com.) recent research also supports this concept.

Additionally, the archaeological record provides some indirect support for the above theory. The Baker collection contains 12 distal fragments of Folsom preforms and 10 have the bevel bias on the face being fluted. During a cursory survey of 6 distal fragments, Tunnel (1999: pers. com.) reported five with the same orientation from the Adair-Steadman material. LeTourneau's (1999: pers. com.) records of the Rio Rancho site captured insightful information on 4 distal fragments. Three of these 4 had the bevel on the face being fluted. This count of 18 of 22 having bevels on the side being fluted allows one to reject the null hypothesis that the position of the bevel is a random orientation. This is definitely statistically significant with an alpha of 0.005.

SUMMARY, CONCLUSIONS, AND QUESTIONS

This paper reported on the in-progress research of computer modeling the fluting of a Folsom point. The software type employed was Finite Element Analysis, which is commonly used to model real world objects under different loading and boundary conditions. FEA is routinely used by design engineers to test their designs before they are constructed. However, this is the first time it has been applied to the Folsom fluting subject.

The FEA software successfully replicated the Pelcin's (1996) experimentally created flakes. It was able to reproduce the two flake types Pelcin identified, and it provided a greater understanding of the characters of each. The two flake types are energy-rich (made with large AOBs) and energy-poor (made with small AOBs). Examples of energy-rich flakes are those made during hard hammer biface thinning or blade manufacture. Energy-poor examples are biface thinning with a billet that produces outré passe flakes and blade making with chest crutch.

When the FEA software was directed toward Folsom fluting, it demonstrated that channel flakes could be made in either of the two energy modes. It indicated that the flake creation mechanics of the two modes were different, and therefore each mode should create channel flakes with unique characteristics. Deep step or hinge terminations are a characteristic of many energy-rich flakes. Many Folsom preform fragments and finished points have this characteristic. This suggests that the Folsom knappers were operating in the energy-rich mode. Is it possible some were operating in the energy-poor mode? If the Folsom people were fluting in both modes can the modes be identified and used to understand group size or dynamics?

The overshot flake is possibly the most common fatal fluting failure. Any software used to model Folsom fluting must be able to replicate this flake. The software used during this research was able to model the overshot flake and it demonstrated that overshots could be created in both the energy-rich and poor modes. The software also provided an insight into the position of the bevel on the distal end of the preform. Moving the bevel bias toward the face being fluted reduces the chances of an overshot failure. The archaeological record suggests the Folsom people were aware of this and were positioning the bevel to reduce overshots. The distal tips in the Folsom collections around the country need to be examined for the position of the bevel in relation to the face being fluted to expand on the small sample presented in this paper.

In conclusion, computer modeling of the mechanics of Folsom fluting with FEA software has been very successful. FEA is powerful tool. It has provided new insights into old observations about the archaeological record and fluting in general. As always happens, new insights raise new questions, such as "did different bands flute in different modes?" In a sense, new questions flow from new technology and old questions tend to stagnate with unchanging technology. Therefore, for archaeology to grow its science it needs to stay on the cutting edge of technology.

ACKNOWLEDGEMENTS

I want to thank Dr. Andrew Pelcin for his Ph.D. research and his eagerness to share it with me. He has been a colleague and constant advisor since the inception of this project. I want to thank Bob Patten for forcing me to investigate the small angles of blow, where I learned that channel flakes can be make in the energy-poor mode. Finally, I want to thank Phil LeTourneau for introducing me to the concept of the distal end beveling.

REFERENCES CITED


NOTES

1 The "web" page address detailing the testing of the FEA model against Pelcin's work is http://www.ele.net/algor/ff_intro.htm.

2 See Note 1.

3 Glass Properties (Richards et. al. 1960:215)

4 As the channel flake develops, it rapidly becomes more flexible in the horizontal direction than the vertical (as view in Figure 4). Therefore, it deflects more horizontally and the horizontal spring consumes a larger percentage of the horizontal force than the vertical spring does of the vertical force. Near the end of the formation of most flakes the horizontal force in the energy-rich mode is almost removed from the channel flake by the horizontal spring. The vertical force remains on the channel flake since there is little deflection in this direction. So the channel flake outruns the indentor in the horizontal direction only. It does not outrun the vertical force.

5 During the early history matching of Pelcin's research, flakes could not be created with the computer model that would match the flakes Pelcin had created in the laboratory. They were always too short and would not run. At that time the need for elasticity (springs) had not been discovered so there was only force applied to the model. This force was actually two forces, a vertical one (compression into the flake) and a horizontal one (bending the flake). It was the horizontal force that was causing the problem. As the flake would grow in length, the horizontal force would gain leverage from the flake length (cantilever beam effect). This leveraging effect of the force would gradually turn the crack toward the dorsal surface of the flake and cause it to terminate too soon. Something was needed to buffer the horizontal force.

Adding springs to the model that reduce the horizontal force as the flake grew and bent away from the core solved the problem. It solved it so well that the crack ran too far. It would not stop. By modeling a number of Pelcin's flakes, it was discoveed that they all stop their vertical travel at the same change in potential energy per length of crack growth. This value was 1E-9 inch-pounds per millimeter. In different words, as the crack propagates it gives up potential energy and it gives it up at lesser and lesser amounts per millimeter of crack extension. At the above stated amount Pelcin's cracks stopped.

Now the problem was that the crack had stopped while its tip was in the body of the core. Something was needed to break the flake out. All that was necessary to do this was remove the springs and, therefore, reapply the full horizontal and vertical forces. Since the flake was already longer than it would ever have grown without the springs, the crack immediately broke to the dorsal surface causing a sudden change in direction of the crack. This is the step or hinge termination.

Additional investigation proved that the flakes first created with the model without the springs were the energy-poor flakes. Slowly the following theory began to develop. When the indentor hits the core and both rebound from the impact. If the location of the blow and the AOB places the core in the energy-rich mode, a crack propagates at a high rate of speed until it exhausts its energy. Then the indentor rebounds from the impact and the full horizontal force is applied and breaks the flake out. Results are a hinge termination. If the location of the blow and AOB places the core in the energy-poor mode, a crack may start but it travels only a fraction of a millimeter and waits for the indentor to rebound and continue propagating the crack. Results are a shorter flake with a gradual rise to the dorsal surface.

6 Perform splitting is considered a result of improper striking of the platform, e.g. a second strike during the follow through by the indentor. They are not similar to the overshot flake that is a natural mechanical phenomenon that results for certain preform morphologies, AOBs, TPTs, etc. Therefore, they are not considered a normal fluting result if the platform is properly struck.

7 In Figures 12c-d the channel flake has a hook shape near the platform. This shape is actually present in the FEA model. The channel flake has moved away from the horizontal force of the indentor and only the vertical force remains. So the channel flake is buckling near the platform where it is the thinnest. The exaggeration of the horizontal 350,000 times makes this look unnatural.