USF Libraries
USF Digital Collections

An examination of a three-dimensional automated firearms evidence comparison system


Material Information

An examination of a three-dimensional automated firearms evidence comparison system
Physical Description:
Carpenter, Natalie G
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:


Subjects / Keywords:
Dissertations, Academic -- Criminology -- Masters -- USF   ( lcsh )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


ABSTRACT: This thesis is an examination of a firearm identification system that creates a three-dimensional image of a bullet in order to record the depth and length of striations occurring along the bullet's surface. Ballistics evidence is an area of forensics in great need of further development. The advent of more sophisticated firearms such as semi-automatic and automatic weapons has increased the need for a matching system that connects bullets found at crime scenes with suspect guns. In the past, control bullets matching ones found at the crime scene have been test fired and then examined by a comparison microscope for similarities with the evidence bullet. The purpose of this thesis is to examine data collected by an emerging system that uses three-dimensional technology by way of a laser and convex mirrors to create a digitized representation of the lands and grooves of a bullet. This representation is a measure of the depth of striations or markings created on the bullet's surface during the firing event. The objective of this thesis is to statistically examine the data collected by this system, which consists of bullets produced by eight different manufacturers. The data for this thesis comes from a pilot study conducted by the creators of a three-dimensional system called SCICLOPS. Variables examined include the maximum and minimum number of striations recorded, the relative position of the bullet (as determined by the six lands and grooves measured by the system), and the manufacturer type. It is hypothesized that there will be differences in the number of striations measured across manufacturer types. Results indicate that manufacturer type may play an important role in how bullets "take" striations or markings during the firing event. Implications for the SCICLOPS system and future research are discussed.
Thesis (M.A.)--University of South Florida, 2004.
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Natalie G. Carpenter.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 67 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001469429
oclc - 55731840
notis - AJR1183
usfldc doi - E14-SFE0000330
usfldc handle - e14.330
System ID:

This item is only available as the following downloads:

Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001469429
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 040524s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000330
b SE
1 100
Carpenter, Natalie G.
3 245
An examination of a three-dimensional automated firearms evidence comparison system
h [electronic resource] /
by Natalie G. Carpenter.
[Tampa, Fla.] :
University of South Florida,
Thesis (M.A.)--University of South Florida, 2004.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 67 pages.
ABSTRACT: This thesis is an examination of a firearm identification system that creates a three-dimensional image of a bullet in order to record the depth and length of striations occurring along the bullet's surface. Ballistics evidence is an area of forensics in great need of further development. The advent of more sophisticated firearms such as semi-automatic and automatic weapons has increased the need for a matching system that connects bullets found at crime scenes with suspect guns. In the past, control bullets matching ones found at the crime scene have been test fired and then examined by a comparison microscope for similarities with the evidence bullet. The purpose of this thesis is to examine data collected by an emerging system that uses three-dimensional technology by way of a laser and convex mirrors to create a digitized representation of the lands and grooves of a bullet. This representation is a measure of the depth of striations or markings created on the bullet's surface during the firing event. The objective of this thesis is to statistically examine the data collected by this system, which consists of bullets produced by eight different manufacturers. The data for this thesis comes from a pilot study conducted by the creators of a three-dimensional system called SCICLOPS. Variables examined include the maximum and minimum number of striations recorded, the relative position of the bullet (as determined by the six lands and grooves measured by the system), and the manufacturer type. It is hypothesized that there will be differences in the number of striations measured across manufacturer types. Results indicate that manufacturer type may play an important role in how bullets "take" striations or markings during the firing event. Implications for the SCICLOPS system and future research are discussed.
Adviser: Tom Mieczkowski
0 690
Dissertations, Academic
x Criminology
t USF Electronic Theses and Dissertations.
4 856


An Examination Of A Three-Dimensional Automated Firearms Evidence Comparison System by Natalie G. Carpenter A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Criminology College of Arts and Sciences University of South Florida Major Professor: Tom Mieczkowski, Ph.D Kim M. Lersch, Ph.D M. Dwayne Smith, Ph.D Date of Approval: April 8, 2004 Keywords: manufacturing,matc hing,statistics,guns,bullets Copyright 2004 Natalie Carpenter


i Table of Contents List of Tables iii List of Figures iv Abstract v Chapter One 1 Introduction 1 Firearms Identification 2 Class Characteristics 3 Individual Characteristics 4 Types of Handguns 4 Firearm Manufacturing Techniques 5 Bullet Manufacturing 7 Summary of Bullet Identification 9 Purpose 10 Chapter Two 11 Literature Review 11 Review of Previous Studies 11 Empirical Studies 11 Using Statistics to Test Matching 12 Testing Consecutively Manufactured Barrels 13 Improvements on Previous Studies 15 Summary 16 Legality of Firearms Evidence 16 Court Decisions 16 Meaning of Expert Testimony 18 Introduction to Previous Identification Systems 18 Comparison Microscope 19 Laser Topography System 19 System Testing 20 Automated Systems 22 Basic Components of an Automated System 22 Examples of Automated Systems 23 A Three-Dimensional Automated System 24 Advantages and Disadvantages of 2D and 3D Systems 24 Components of the SCICLOPS System 25 Summary 27 Hypotheses 28


ii Chapter Three 30 Methodology and Data 30 Methodology 30 Data 30 Variables 32 Analysis Strategy 33 Descriptive Statistics 33 Normality Tests 34 Significance Tests 34 Chapter Four 35 Results 35 Descriptive Statistics for Each Bullet Distribution 35 Normality Tests for Each Bullet Manufacturer 36 Significance Tests Across Manufacturers 39 ANOVA 39 Post Hoc Tukey and Homogenous Subsets Tests 40 Differences in Means Within Each Manufacturer 44 Chapter Five 46 Discussion 46 Differences in Striations Measured by Bullet Type 46 Differences Within Manufacturer 49 Limitations 51 Conclusion 52 Future Areas of Study 52 References 54 Bibliography 57 Appendices 58 Appendix A: Histograms and Boxplot s by Orientation and Manufacturer 58


iii List of Tables Table 1 Characteristics of a Firearm 3 Table 2 Descriptions of Bullets Used in Analysis 31 Table 3 Summary of Variable Definitions and Coding 33 Table 4 Descriptive Statistics for Average Weighted Striations 35 Table 5 Normality Tests by Bullet Type for Incorrect Orientation 38 Table 6 Normality Tests by Bullet Type for Correct Orientation 39 Table 7 Tukey Test Between Manufacturers for Correct Orientation 40 Table 8 Tukey Test Between Manufacturers for Incorrect Orientation 41 Table 9 Homogenous Subsets for All Manufacturer Types in Incorrect Orientation 42 Table 10 Homogenous Subsets for All Manufacturers in Correct Orientation 43 Table 11 ANOVA Test for Differences Within Manufacturers 44


iv List of Figures Figure 1. Histogram of Incorrect Orientations for all Bullet Types 37 Figure 2. Histogram of Correct Orientation for all Bullet Types 37


v An Examination Of A Three-Dimensional Automated Firearms Evidence Comparison System Natalie G. Carpenter ABSTRACT This thesis is an examination of a firearm identification system that creates a threedimensional image of a bullet in order to record the depth and length of striations occurring along the bullets surface. Ballistics evidence is an area of forensics in great need of further development. The advent of more sophisticated firearms such as semiautomatic and automatic weapons has increased the need for a matching system that connects bullets found at crime scenes with suspect guns. In the past, control bullets matching ones found at the crime scene have been test fired and then examined by a comparison microscope for similarities with the evidence bullet. The purpose of this thesis is to examine data collected by an emerging system that uses three-dimensional technology by way of a laser and convex mirrors to create a digitized representation of the lands and grooves of a bullet. This representation is a measure of the depth of striations or markings created on the bullets surface during the firing event. The objective of this thesis is to statistically examine the data collected by this system, which consists of bullets produced by eight different manufacturers. The data for this thesis comes from a pilot study conducted by the creators of a threedimensional system called SCICLOPS. Variables examined include the maximum and minimum number of striations recorded, the relative position of the bullet (as determined by the six lands and grooves measured by the system), and the manufacturer type. It is hypothesized that there will be differences in the number of striations measured across


vi manufacturer types. Results indicate that manufacturer type may play an important role in how bullets take striations or markings during the firing event. Implications for the SCICLOPS system and future research are discussed.


1 Chapter One Introduction The term ballistics refers to the stu dy of the motion of a projectile. There are three types of ballistics that are usually studied internal, external, and terminal. Internal ballistics involves the study of the projectile within the firearm and includes the areas of chamber configuration, chamber pressure, and rifling. Exterior bal listics concerns the projectile after it leaves the firearm, i.e. velocity and trajector y. Finally, terminal ballistics concerns the study of the effects of the projec tile on a target. In the Handbook of Forensic Science the Federal Bureau of Investig ation (1981) defines firearms identification as the study by which a bullet, cartridge case, or shotshell casing may be identified as having been fired by a partic ular weapon to the exclusion of all other weapons(p.52). Firearms themselves have had a long, illustrious, and documented history, while the first written reference to th e subject of firearms identification has been recorded as occurring in 1900 with Halls The Missile and the Weapon in the Buffalo Medical Journal It was not until the 1920s, however that the topic gained attention. Calvin Goddard, often credited as the father of firearms identification, was responsible for much of the early work on the subject during his examinat ion of the various kinds of firearms and bullets at his Scientific Crime Detection Laboratory in Chicago. Today, the area of firearm identification c ontains within itself a huge quantity of information. The advent of semiautomatic and automatic weapons calls for a new


2 technology in identification. One system to emerge has been a threedimensional automated firearm identification system. This and other identification systems will be explored in this thesis, along with the hist ory of firearm identification and a breakdown of the parts and manufacturing of firearms. Firearms Identification Firearms identification requires kno wledge of weapons and ammunition. Giannelli (1991) lists rifles, handguns, and s hotguns as the three types of firearms typically used for examination. Firearms can be divided further into smooth bores and rifled arms. Smooth bores are firearms in whic h the bore (inside of the barrel) is perfectly smooth from end to end. A rifled arm has a longitudinal cut with a number of parallel spiral grooves. The surfaces between the grooves are called lands. The lands and grooves twist in either a right-hand or left-hand direction. Manuf acturers specify the number of lands and grooves, the direction of twist, th e angle of twist (pitch), the depth of the grooves, and the width of the lands and gr ooves. Shotguns fall under the smooth bores category, while handguns and rifles are considered rifled arms. Some common firearm terms include bore and caliber. Bore can be used to describe the diameter of the interior of a weapons barrel (Territo, 2000, p.106). In a handgun or rifle, the bore is usually m easured between two opposing lands (ridges). Caliber refers to the diameter of the bullet intended for use in the firearm and is usually expressed in either hundredths or thousandths of an inch (.22, .45 caliber) or millimeters (7.62 mm). The bullet is usually larger than the diameter of the bore, so that the lands grip it as it passes through the barrel. This causes the bullet to rota te, usually in a right-


3 hand direction. This movement creates highly individualized striati ons on the bullet as well as increasing the accuracy. Because the lands bite into the bullet surface, the land and groove impressions are imprinted on the bullet and play an important role in firearms identification. Firearms identification is concerned with two types of characteristics of a firearm: class and individual. Table 1 lists examples of each type of characteristic. Table 1. Characteristics of a Firearm Characteristic Definition Examples Class Characteristics dealing with type and manufacturer Caliber, number of lands and grooves Individual Characteristics dealing with actual firearm itself Barrel deformities, number of striations created during firing Class Characteristics The class characteristics of a firearm include its caliber and rifling specifications: (1) the land and groove diameters; (2) the direction of rifling (right or left twist); (3) the number of lands and grooves; (4) the width of the lands and grooves; (5) the degree of the rifling twist; and (6) th e depth of the grooves. In firearm identification, if the class characteristics do not match, the firearm could not have fired the bullet. Also, if the bullet is recovered before the firearm, the class ch aracteristics could provi de information about the type of firearm that could have fire d the bullet. Thus, identifying the class characteristics of a firearm is useful in matching a gun to a bullet. However, the class characteristics, while useful in determini ng what brand of gun was used, are not helpful in identifying a specific gun. No manufacturi ng process produces one hundred percent identical guns one after another. The rifling process causes unique striations or markings


4 on each gun produced. This can be due to seve ral reasons, but no matter the reason, class characteristics cannot focus attention on one specific gun. The individual characteristics of the gun are the most important when matching a bullet to a gun. Individual Characteristics Once a firearm and an evidence bullet have been matched on cl ass characteristics, a positive identification can be made as to what type of gun a bullet was fired from. But it takes matching the individual characteristics of a gun to a bullet to really be positive that one certain gun was the only gun that could have fired that bullet. Barrels are machined during the manufactur ing process, and any imperfections in the machine are imprinted on the bore. Subseque nt use of a firearm adds more individual markings, such as erosion caused by the fric tion of the bullets pa ssing through the bore or corrosion caused by moisture (rust). As stated previously, thes e individual markings can distinguish one gun and maybe even one bullet of the same type from another. The ability to perform bullet-to-bullet comparisons ba sed on microscopic surface features is therefore at the core of fore nsic firearms identification. The ability to say something such as, Of all of the 9mm revolvers in the world, this is the only one that could have fired this specific bullet would allow for stronger evidence in shooting cases. The question surrounding the issue of identification is whether it is even possible to distinguish two guns or two bullets based on the microscopic features. Types of Handguns Most handguns can be divided into two types, revolvers and semiautomatic pistols. One major difference between the two is that the cartridge case is automatically


5 ejected when a semiautomatic pistol is fired. Revolvers have a cylindrical magazine that rotates behind the barrel, with the cylinder holding around five to nine cartridges, each within a separate chamber. Semiautomatic pistols do not have cylinders; instead, the cartridges are contained within a vertical magazine, which is typically loaded into the grip of the pistol. Rifle and handgun cartri dges (also known as amm unition) consist of the projectile (bullet), case, pr opellant (powder), and primer. The primer contains a small amount of explosive mixture that de tonates when struck by the firing pin. This detonation incites the ignition of the propellan t. Modern propellant is smokeless powder, either single-base (nitrocellu lose) or double-base (nitroce llulose and nitroglycerin). Bullets are generally composed of lead a nd small amounts of other elements, known as hardeners. These bullets may be completely covered with another metal (jacketed,) or only partially covered (semi-jack eted). Bullets may also have different shapes, such as flat base, hollow base, round nose, flat nose, or hollow point. Shotguns, as previously mentioned, do not have lands and grooves. Their shells consist of a case, primer, prope llant, projectiles, and wadding. Wadding keeps the powder and the pellets in positi on inside the shell and may be paper or plastic material. The projectiles are generally spherical balls (pellets). Firearm Manufacturing T echniques There are several different methods of manufacturing firearms. The methods presented here include hook cutting, broach ing, buttoning, mandrel, and drilling. Each of these methods involves a rifling process, in which the barrels inner surface is impressed with spiral grooves. The spiral grooves are important during the firing process, because


6 they guide the bullet through the barrel, givi ng it a rapid spin and, by that, a straight trajectory. The hook cutting method was prevalent prior to 1940. In this method, barrels are rifled by having one or two grooves at a time cut into the surface with steel hook cutters. The cutting tool is rotated as it passes down the barrel to give the grooves direction (to the left or right). The broach cutting method involves a series of concentric steel rings (known as a broach), with the size of the ring incr easing slightly down the line. The broach simultaneously cuts all of the grooves into the barrel at the requi red depth as it passes through the barrel. As in the hook cutting method, th e rotation of the broach in the barrel gives a direction and rate of twist to the grooves. In the button process, a steel plug or bu tton impressed with the desired number of grooves is forced under extremely high pre ssures through the barrel Only a single pass is necessary to compress the metal and crea te lands and grooves on the barrel walls. The rotation of the button, as with the other met hods, gives the grooves a direction and rate of twist. In the mandrel rifling process, a rod of hardened steel is molded and formed so that the shape is the reverse impression of the rifling it is intended to produce. This rod is inserted into a slightly oversized bore, and the barrel is compressed with hammering or heavy rollers into the mandrels form. The rod is then removed, leaving the finished product with impressions on the inside of the barrel.


7 Lastly, the drilling process involves a ba rrel being produced from a solid bar of steel that has been hollowed out by drilling. This drilling leaves microscopic marks on the barrels inner surface. The drilling process is a mo re modern technique, but imperfections in the manufacturing equipment are still a cause of microscopic marks on the inside of the barrel. These manufacturing processes determine the class characteristics of firearms. Since no two manufacturers use exactly the same method or equipment, firearms can be distinguished by the manufacturi ng process used to create th e barrel. One can tell a Luger from a Sig because of the class characteristic s associated with each firearm type. This has become important in the area of fire arm identification due to the number of manufacturers and the large number of firearms in use. A preliminary step in matching a suspect bullet to a suspect firearm is checki ng the suspect gun for the class characteristics that could distinguish that gun as a certain make. Bullet Manufacturing Similar to firearms, bullet manufacturi ng has a myriad of types and methods. Williams (1980), in his book Practical Handgun Ballistics tries to break down the major categories of bullets. He categorizes bullets into three types soft lead bullets, hard lead bullets, and jacketed bullets. Williams describes two processes for manufacturing bullets the cast lead process and the commercial swag ed lead process. For a cast lead bullet, manufacturing is simply a process of me lting lead and pouring a small amount into a mold, allowing it to cool and harden, and th en removing that small amount from the mold. If done correctly, the bullet resembles its fi nal shape. But it must be noted that this


8 is not the final bullet, as the product formed from the mold is somewhat deformed. It takes the lubricating and shaping process to create the final product. This process involves running the bullet th rough a sizing die that is prec isely the same size as the barrel the bullet is in tended for (i.e., .357 Magnum). This die is involved in the final forming of the bullet. Lubrication involves a sort-of cleaning of the bullet in which the surface of the bullet is made smooth. If lubric ating is not done, lead scrapings from the bullet would coat the barrel and clog it. Lubricating help s prevent lead buildup on the barrel of the gun and helps to lengthen the life of the barrel. A special aspect of cast lead bullet manufacturing to cons ider is the nose type. Williams (1980) lists four types of cast lead bullet noses: wadcutter, hollow-point, round nose, and Keith type. The nose of the bullet is important, as it is the first part of the bullet to come into contact with the target. Older bullet nose types, such as the wadcutter and round-nose, were found to either not work in high velocity barrels (wadcutter) or to have such problems as excessive penetration and deflection upon hitting hard surfaces (round-nose). The hollow-point bullet nose wa s an improvement, as it expanded shortly after impact (causing more damage to the targ et). The Keith type was formulated to bring together the best aspects of the hollow point and the round no se. When the bullet strikes, the nose would expand like the hollow point, but the heavy, solid middle part of the bullet would push forward and penetrat e the target w ith greater force (lik e the round nose). The problem with lead cast bullets was found to be that they did not stand up to the high temperatures created in a hi gh-velocity barrel (m ost modern guns). Thus came the advent of the jacketed bullet.


9 The premise behind the creation of the jacketed bullet is that encasing the lead bullet within a gliding metal cover or jacket would allow the bullet to be fired through high-velocity gun barrels without melting or deforming. A long, lengthy process ensued the advent of the jacketed bullet, due to th e question of how to k eep the jacket on the bullet during the firing process, as it was know n to blow off when leaving the barrel. It was found that crimping (pressing with a mach ine) very long jackets over the bullet kept the jacket stable so it would not fall off. Summary of Bullet Identification The procedure normally used in bullet id entification involves a comparison of the evidence bullet and a test bu llet fired from the weapon. Th e test bullets are usually obtained by firing a firearm into a recovery box, a bullet trap (filled with cotton), or a recovery tank (filled with water). The two bullets are then compared by means of a comparison microscope, which permits a split-screen view. This allows for visual identification of striations and other marks. The firing of a bullet through a barrel is thought to create unique markings on the bullet s surface. The question then evolves as to whether it is then possible to identify a bulle t by its unique characteristics as coming from a specific gun. As stated before, it is thought that the unique characte ristics of a bullets surface come into play during the manufacturi ng process. Therefore, researchers have examined bullets and firearms created by different manufacturers in order to find similarities or differences in striations created during the firing event. Several studies have been conducted on th is question starting back in the early years of firearm manufacturing, and they have emerged with mixed results.


10 Purpose The purpose of this thesis is to examine the next step in firearm identificationa threedimensional imaging system that digitizes the ridges and grooves created on a bullets surface during the firing mechanism. The research questions surrounding this issue include the following: whether bullets can be differentiated by manufacturer; whether all bullets of a single ma nufacturer will match each other; whether the imaging system reads differences in bullets.


11 Chapter Two Literature Review The ability to compare bullets by exam ining microscopic striations on each bullets surface is at the hear t of ballistics assessment. As stated before, microscopic striations are formed on a bu llets surface during the firing se quence. Some causes of this include structural imperfections of the fir earm or pressure created during the firing sequence. It has therefore b een thought possible to match one bullet to another by firing both bullets from the same firearm. Studies investigating this possibility have emerged with mixed results. Review of Previous Studies Nichols (1997), in his e xhaustive review of firearm and toolmark identification literature, examined thirty-four articles dating from 1949 to the present. Empirical studies conducted on bullets and casings fired through the same weapons have made up the majority of research. Nichols reports the earliest empirical study on firearm identification to have been conducted by Churchman in 1949. Empirical Studies Churchman (1949) analyzed characteristics typical of the Cooey .22 caliber rifle barrel. He emphasized the importance of knowing the origin of markings on bullets before one could utilize them for th e purposes of unequivocal identification. The Cooey rifle was manufactured using the broachi ng technique, which Churchman believed was


12 responsible for producing sub-cl ass characteristics on the bulle ts (striations at the edges of the land impressions). He examined test -fired bullets from three consecutively broached rifle barrels. He found that the br oach characteristics pe rsisted from barrel to barrel. However, he also found individual charac teristics of each rifl e that did not carry over to the other two. Using Statistics to Test Matching Biasotti (1959) conducted a statistical ev aluation of the individuality of bullets fired from different firearms. Using a to tal of twenty-four .38 SPL Smith & Wesson revolvers in the comparison, Biasotti gather ed different combinations of bullets, land impressions, and groove impressions. Sixteen of the revolvers had previously been fired, while the last eight were new. The sixteen used revolvers were gr ouped together, and the test bullets fired from these revolvers were compared amongst the other bullets fired by the same revolver and the test bullets fired by the other fifteen. Groups II and III consisted of the eight new revolvers. Group II contained the same bullet types as Group I (158 grain solid lead bullets) while Group III fired 158 grain jacketed bullets. Biasotti then evaluated the different impression combinations for percentage of matching striations and consecutiveness. In order to do the analysis, Biasotti developed terms for the striations. A line was defined as an en graving or striation appearing on a bullet as a result of being engraved by the individual irregular ities of characteristics of the barrel, plus any foreign material present in the barrel capable of engraving th e bullet (p. 36). So each line was an individual characteristic. Consecutiveness was defined as the compounding of a number of individual character istics (p. 36). This would be defined as


13 class characteristics, in that Biasotti wanted to see if individual characteristics carried over to the other guns of the same type and manufacture. Any consecutiveness would mean that individual characteristics were not actually unique. Biasotti thus evaluated both quantity (objective feature) and quality (subjective feature). He found that the average percentage of matching lines in jacketed bullets fired from the same gun was 21-24%, and 15-20% matching striations on land or groove impressions between bullets fired from different weapons. For consecutiveness, Bias otti found no more than three consecutive matching striations for lead bullets fired fr om different weapons and no more than four for the jacketed bullets. Testing Consecutively Manufactured Barrels Lutz (1970) published one of the first st udies on the correspondence of markings on bullets test fired from consecutively rifl ed barrels, meaning that the barrels were manufactured one right after the other. Lutz fired a series of jack eted and lead bullets through each of two unused .38 SPL barrels. He then fired a second se t of bullets through each barrel and had them coded. Firearms exam iners were then asked to compare the first set of bullets (test se t) to the second, coded set. The re sults indicated that the examiners were able to easily identify th e barrel of origin for each of the bullets and that there were many dissimilarities of land impressions from each barrel. Skolrood (1975) conducted a study similar to that of Churchman. He performed a series of comparisons on bullets fired from three new, consecutively broached, .22 caliber Winchester rifle barrels. He found that compar isons of bullets fired from the same rifle yielded more persistent characteristics than comparisons on bullets fired from different


14 rifles. Thus, bullets fired from a specific gun had a higher matching rate than bullets fired from other guns of the same type and manufacturer. Freeman (1978) conducted a study on thr ee consecutively rifled, Heckler & Koch, 9-mm Luger caliber, polygonally rifled barrels. He found that each barrel was distinctly individual, and that, although th e first two barrels could be easily inter-compared, the third barrel yielded poorly marked test bullets. Thus, even consecutively rifled barrels contained individual characteris tics, even though they were manufactured one after the other. Murdock (1981) empirically assessed the individuality of button-rifled barrels. In this study, he discussed the various forms of early cut-rifling methods and the idea that these methods left sub-class features on barre ls. He also discussed the newer methods of rifling that did not involve th e removal of any metal, which is the opposite of the earlier methods. Assessing the individuality of .22 caliber barrels, he found no continuity of subclass characteristics in the bullets fired from each of the three barre ls. In a similar study conducted in four Shilen DGA barrels, Hall (19 83) found that test-fir ed shots closer in firing sequence showed more similarity than te st-fired shots further apart in the sequence. He was able to conclude that, with bullets closely related in the firing sequence the dissimilarity of marks created by any two differe nt barrels is significantly greater than the dissimilarity seen on bullet pairs that are from the same barrel (p. 45). In contrast to the previous studies, Matty (1985) conducted comparisons on three revolver barrels all cut from the same section of rifled tube. He had observed that the buttons used to rifle the barrels did acquire some damage and wanted to see if the


15 damage was transferred to the bore surface. Matty did observe longitudinal striations on the groove impressions caused by button imperfections, of which a few persisted along the length of all three barrels. He found that there was a settling-in period during which test fired bullets from the same barrel could not be identified to each other. This was important because of the questi on of how similarity between bullets could be proven with newly manufactured guns. Matty also found th at, after the settlingin period, comparisons of bullets fired from different barrels proved inconclusive for groove impressions and showed no consistency for land impressions. Improvements on Previous Studies Brundage (1992) conducted a replication of Lutzs (1970) study, with some significant improvements. He provided a pair of test-fired bullets from ten consecutively rifled Ruger barrels to 30 laboratories acr oss the country, along with fifteen unknowns. All of the laboratories properly associated the unknowns with the barrel from which they were fired. This was an improvement over Lu tzs study in that the examiners were not provided any information regarding barrel or test manufacture. Lastly, Brown and Bryant (1995) compared barrels from multi-barreled derringers in an attempt to determine whether the barrels in these weapons may have been consecutively manufactured. Brown and Bryant indicated that, a majo r contributor to the individual bullet striation from the button rifled barrels is certainly the compressed reamer marks that appear very prominently in the casts of the lands and grooves (p. 256). This meant that the marks transferred as individual markings to the surface of the


16 bullets and would not be considered class ch aracteristics but indivi dual characteristics that could show consecutiveness among bullets fired from a single gun. Summary As stated previously, the literature has shown mixed results for the comparison of bullets fired by identical or dissimilar fir earms. The lack of consistent methodology and scientific experimentation in these studies has shown the need for more advanced analyses of firearms and bullets. The importan ce of this experimentation lies in the area of forensics. The question of whether a sing le bullet could be matched to a single gun, if answered, could provide a new direction in shooting cases. Suspects could be tied to a shooting by evidence concerning wh ether their gun is the only one that could have fired a certain bullet and created th e unique individual characte ristics found on the bullet. Creating such a system is only one step in the process. Another impor tant area of firearms identification lies in the legal usefulness of this kind of information. A policeman may be able to match a bullet to a gun and therefore a suspect, but the courts must decide the admissibility of this sort of evidence. Legality of Firearms Evidence Since the beginning of firearms identification, the courts have had to make decisions of the permissibility of this sort of information as evidence. Court Decisions on Firearm Identification Evidence Inbau (1999) conducted a review of important court deci sions regarding firearms identification. Dean v. Commonwealth (1879) was found to be the first case in which an appellate court approved of testimony regarding the simila rity between test bullets and


17 bullets used in a crime. In the 1881 case of State v. Smith the court refused the defendants request to permit an expert to examine and experiment with the evidence pistols to determine which was possibly the one to have fired the suspect bullet. Inbau (1999) stated that this decision was important only for the reason that it apparently represents an early attempt at judicial recognition of the science of firearms identification. The matching of suspect a nd test bullets was first approved by an appellate court in the 1902 decision of Commonwealth v. Best. The evidence presented included photographs of a test bullet having been pushed through the defendants rifle barrel. The court agreed with the evidence, stating that the information provided by the expert witness concerning how a test bullet would be marked during firing was a question of much importance to the case. Laney v. United States (1923) was a federal case that involved firearm identification in its decisi on, in which it was considered admissible for an expert to testify on the matching between a bullet and a pistol. Within the next two decades, the cases of State v. Boccadoro (1929) Galenis v. State (1929) and People v. Beitzel (1929) all affirmed the admissibility of firearm identification testimony. Evans v. Commonwealth (1929) was considered to give the first exhaustive opinion on firearms identification as a science. People v. Fiorita (1930) included in its opinion a guideline against incompetent firearms e xpert testimony, stating that wh ile the science of ballistics is now a well-recognized science both in this country and abroad, testimony based upon it should be admitted with the greatest care. No witness should be permitted to testify regarding the identification of firearms and bullets by the use if this science unless the witness has clearly shown that he is qualified to give such testimony.


18 Meaning of Expert Testimony Inbau (1999) described the kinds of expert testimony required in court outside of bullet/gun matching. He listed the distance a nd direction at which a shot is fired, similarity in the size and wei ght of bullets, proof that a bul let was fired from a weapon of a certain caliber, proof that wounds were cause d by a specific type of firearm, and to prove that a suspected gun was recently fired as others of interest to courts in the area of firearms identification. More recent cases have reaffirm ed the precedents set by the former courts in admitting evidence of bullet, cartridge case, and even shot shell identifications. It seems that the admissibility of firearms identification evidence has been well-established by the court system. Yet it still remains to be seen as to how far in the future this admissibility will last, for as manufacturing techniques become more sophisticated, differentiation be tween bullet types may not be possible. Let us hope that as manufacturing techniques become more sophisticated, so too will identification systems. As can be seen below, this may indeed be the future trend. Introduction to Previous Identification Systems The need for a standardized, highly accura te firearms identification system has been shown throughout the history of firearms identification. During th e early parts of the twentieth century, a magnifying glass was the to ol most often used in the examination of firearms and bullets. Police or other firearm experts would make the decision of whether a bullet and gun matched by visually examini ng the two. This method did not last long, for the advent of the comparison microscope made possible photogra phs of two bullets showing similarities and differences.


19 Comparison Microscope Inbau (1999) detailed the wo rkings of this system of identification. The comparison microscope consists of two ordi nary microscopes arranged in a way that images passing through both are brought t ogether in one eye-piece midway between them. Each bullet (a test bu llet and the suspect bullet) is placed under each lens, and, by properly focusing the instrument and placing the bullets in the sa me orientation, the microscope transmits the fused picture of th e two bullets. The two pictures were merged together as one. If the two bullets were fi red from the same weapon, there would be very little difference between them in the way of markings and striations. This was an innovative technique in its da y, as it was possible to make a visual inspection of two bullets at the same time. Unfortunately, this system contained flaws in the accuracy of the picture projected and the ability of an expert to make a decision concerning the matching of a gun and bullet. More sophisticated and faster paced techniques were needed to accumulate the ever-growing number of comparisons to make The laser topography system is one such innovative technique. Laser Topography System A study published by De Kinder, Prevot, Pirlot, and Nys (1998) introduced a new technology for firearms identification laser topography. The authors stated the problems with the previous system of comparison micr oscopy to be difference s in light intensity (global or for different regions of the object under study), the surface material (nickel or


20 copper), type of light used (temperature of the light source), and angle of incidence of the light (how light hit an object in order to be reflected b ack). Laser topography was an improvement over comparison microscopy because it accurately measured the topography of the surface. It did this by focusi ng an infrared laser on the objects surface. The reflected light was collected by the same lens and detected by a diode array, which means that light was reflected onto a surface, and a laser keep track of where the light went and what part of the surface the light wa s measuring. This signal was used to correct the position of the focusing lens in such a way as to keep the focus of the laser spot on the surface; thereby keeping the position of the lens correspond ing to the distance to the surface relative to a common reference plan e. This compensated for any sliding or movement on the part of the bullet. The ra nge was 1 micrometer to 0.1micrometers, the highest difference in height that c ould be measured by the apparatus. System Testing The equipment was tested in the following areas: static noise, positioning accuracy, reproducibility, and correctness of the measurement. The testing of static noise resulted in the detection of a substantial backlash, leading to the development that surfaces had to be measured while scanning in the same direction. This meant that the data received by the laser was not being sent, because there was too much for the laser to filter through to find the signal. This doubled the measurement time. The positioning accuracy of the rotational stage was veri fied, as well as the reproducibility and correctness of the measurement. This showed where to place the bullet so that it would be scanned correctly. Optimal scanning speed was found to be 0.5 1 mm/s.


21 As for the testing of how the system act ually measured bullets, tests were also conducted on striation marks on 9 mm Para bullets to indicate whether the topography system could compete with the comparison mi croscope. Only one striation mark on the bullets was the focus, in order to verify th e origin of the striation. This was thought to lessen any chance of comparing bullets on diffe rent sides from each other. The following bullets were studied: an unused bullet (for reloading purposes ), a bullet orig inating from an unfired round, fired bullets of different type (lead or jacketed), and fired bullets caught by different traps (water or cotton wool). This would enable the experimenters to differentiate between fabrication marks, stri ations made by the barrel, and marks left during the bullet recovery process. The bullets were fired with a Fabrique Nationale High Power pistol, resulting in six grooves with a right-hand twist. The measurements were then made by the topography system on one stri ation mark. A correction for the curvature of the surface had to be performed. Results showed that the ja cketed bullets bore no characteristic marks from the fabrication process apart from the normal circle created during the firing process. However, lead bul lets were found to have fabrication marks. There was no evident difference between bullets recovered in the water tank and in cotton wool. The topography measurement of the one st riation in the bullets was thought to be indicative of the systems success in measuring the same phenomena as the comparison microscope and proof that the grazing a ngle illumination (laser making slow passes across whole object) had a very high sens itivity for detecting small topographical differences. In scanning the su rfaces of bullets, th e laser topography system was forced to superimpose the obtained profile on a slowly varying sinusoid. The experimenters found


22 it evident that the striation marks found carried characteri stic information; however, they took this to mean that less measurements were necessary to extract the characteristic information. Not all lands and grooves were measured, and yet the recording time was still higher than the comparison micros cope method. Although laser topography was a step in the right direction, further advancemen ts were necessary, especially in the areas of scanning and measuring the entire bullet. Automated Systems The continuous evolution of smaller, more powerful computers since the 1990s has heralded the arrival of a powerful screen ing tool for firearm identification experts. Automated search and retrieva l systems have the objectiv e of enabling the comparison of evidence and control bullets, therefore transforming forens ic ballistic analysis from an evidence verification tool into a crime-fighting tool (Bachrach, 2002, p. 1). Basic Components of an Automated System The two basic components of an automa ted system are the acquisition and the correlation components. The acquisition component involves the capturing of data and encoding it in order to make it analyzable. Data that has been enc oded and processed is referred to as normalized data. The corre lation component, however, is responsible for making sense of the normalized data, through co mparing the sets of data and organizing the results for the users in spection. Bachrach (2002) specifi es the correlation component as including all the software elements necessary to: a.) Evaluate the degree of similarity between two sets of normalized data


23 b.) If more than tw o bullets are involved in a co mparison, to organize the results of a set of comparisons in some convenient way, and c.) To provide the user with tools to verify the results obtained by the correlation algorithms. Examples of Automated Systems Two major automated systems have alr eady been developed: the Integrated Ballistics Identification System (IBIS) and DRUGFIRE. These two systems have many points in common, such as the capability of acquiring data from bullets and cartridge cases, storing this data in a database, and us ing the database to pe rform comparisons on a given bullet. The most important area of comparison between the two systems is the use of a two-dimensional representation of the surface of the specimen. IBIS processes digital microscopic images of identifying features found on both expended (already fired) bullets and cartridge casings. DRUGFIRE emphasizes the examination of unique markings on the cartridge cases expended by the weapon. The data capture processes in both systems use a source of light directed at the bulle t or cartridge casing s surface to reflect striations, land impressions, and groove impr essions for a camera to record. Bachrach (2002) notes that, when using light as a sour ce, the incident light angle and the camera view angle cannot be the same in order to ob tain a pattern of dark -and-bright reflections of the bullets surface. This accounts for the method of side lighting in two-dimensional imaging, and thereby makes this method an indi rect measurement of the bullets surface. Bachrach introduces a three-dimensional process believed to improve upon the twodimensional systems.


24 A Three-Dimensional Automated System The SCICLOPS system, based on the use of a three-dimensional characterization of the bullets surface, has as its source conf ocal sensors, which operate by projecting a laser beam through a lens onto the surface of th e object and detecting the reflection of the laser with the same lens. This is an imp rovement over the laser topography technique proposed by De Kinder et al. ( 1998) in that the sensor contin uously displaces the lens in order to maintain the laser and allow for an accurate imaging of the entire bullet. Unlike the IBIS and DRUGFIRE programs, the angle of incidence and the a ngle of reflection of the laser beam are the same, so there is no side-lighting. The data acquired is therefore the distance between the surface features and an imaginary pl ane, as the measurement is made along a direction perpendicular to the surface. Advantages and Disadvantages of 2D and 3D Systems Some disadvantages of the two-dimens ional system include the robustness and discontinuity of the data. Bachrach (2002) states a significant problem associated with 2D data capture to be the fact that the transformation relating the light incident on the bullets surface and the light reflected by it depends not only on the striations found on the bullets surface, but also on a number of i ndependent parameters such as the light incident angle, the camera view angle, varia tions on the reflectivity of the bullet surface, light intensity, accurate bullet orientation, etcimplying that the captured data are also dependent on these parameters (p. 3). Another problem is the phenomenon of shadowing, in which some of the smaller features can be shadowed by larger featur es. This shadowing could cause inaccurate


25 reflections of the captured data. This problem is not unique to two-dimensional systems, as the SCICLOPS systems laser beam re quires an unobstructed conical region to properly operate. This limits the steepness th at the confocal sensors can measure. The acquisition speed of two-dimensional systems is significantly faster than the threedimensional SCICLOPS system, allowing examin ers to make decisions more quickly. In comparing the DRUGFIRE system with SCIC LOPS, Bachrach found that the SCICLOPS system created a clear definition of the tran sitions between land and groove impressions, whereas the same boundary was not as well-defined by the DRUGFIRE system. Components of the SCICLOPS System As with the study by De Kinder et al (1998), the SCICLOPS system has a measurement resolution of 0.1 micrometers in depth and 1 micrometer in lateral resolution, thought to be signifi cant enough to capture the most significant elements of the surface data. Experimentation showed the final configuration of the acquisition unit to be on the order of 1 micrometer, as it wa s limited by sensor and mechanical vibration noise. The digitization process involves taking cr oss-sections of the bullet and measuring land and groove impressions, with a suffici ent number of cross-sections giving a complete description of the bullet as a thre e-dimensional object. The geometric region defined by the cross section is approximately an elliptical, because of tilt. The data normalization process of SCICLO PS then conceptually consis ts of two steps: estimation of the ellipse defined by the geometric lo cation of the land impressions (the crosssection) identified in the acquired data and the projection of the acquired data onto the estimated ellipse. The second step corrects for any deformation of the bullet whether in


26 the structure or the acquisition process. For the correlation component, SCICLOPS receives as an input the normalized data of two bullets for matching purposes. The output returns the following information: relative orientation at which the two bullets are most similar and a similarity measure (0 = no sim ilarity up to 1 = identical). The similarity measure used is the correlation function. Th is is a normalized (maximum value is 1) quantification of the degree of similarity between two bullets. A macro and micro correlation are computed while comparing the two bullets in different relative orientations. The macro correlati on is obtained at the orientation in which the two bullets are most similar, while the micro correlation is taken at th e most dissimilar orientation. The Composite Correlation is the geometric av erage of the macro and micro correlations and an overall measure of similarity. A preliminary evaluation conducted by the researchers showed the system to produce reliable characterizations of a bullet surface and to successfully identify similarities between bullets fired by the sa me gun. Problems of the SCICLOPS system as noted by the researchers include the use of only pristine bullets in the creation and evaluation of the SCICLOPS system creating a need fo r acquisition and correlation algorithms for damaged bullets, statistical me thodologies to quantify the performance of automated systems, the need for determining how likely it is that the said bullet was fired by the same gun as the evidence bullet, and a consensus on which is the best location on the bullets surface to acquire the data.


27 Summary The SCICLOPS system is thought to represen t the next generation in firearm identification with the creation of a threedimensional image of a bullet that would accurately represent all striati ons and impressions on the bullets surface. The history of forensics and firearm identification in partic ular has shown the need for a comprehensive system of comparing evidence bullets with test bullets in order to match a suspect gun to a shooting. The creation of striations and impressions on a bullet s surface during the firing process allows for an examination of whether the striations and impressions are consistent among bullets fired by the same gun. The advent of computers has allowed for faster, more comprehensive processing of striations and impressions on a bullets surface than the original comparison microscope did. Th e large quantity of guns being used in the United States and in shootings shows the n eed for a database of firearm and bullet characteristics. The SCICLOPS system allows for a three-dimensional image of a bullet created by taking cross-sections of the bulle ts surface and representing them on a plane in space. This system allows for the comparison of two bullets, just as the comparison microscope, but the SCICLOPS system computes a correlation functi on detailing how the two bullets compare mathematically. This thesis focuses on the SCICLOPS system and the correlation functions computed by the imaging process. When bullets are matched perfectly, there will be a corre lation of 1.0. As there are many types and manufacturers of bullets, it remains to be seen whether bullet types are affected by or themselves affect the striations and impressions created on bullets during the firing process.


28 Hypotheses Based on the existing literature and de scription of the SC ICLOPS system, two hypotheses about ballistics matching can be drawn. However, this project focuses on one area of ballistics ammunition. Given that there are many manufacturers of bullets, it seems appropriate to hypothesize that there wi ll be differences in the amount and quality of striations and impressions made during the firing event. The reason behind this concerns the differences in the quality and manufacturing processes of ammunition today. Some manufacturers have sophisticated high-t ech processes that cr eate identical bullets, while other manufacturers may not have such high standards. In other words, bullet manufacturers will make a difference in the abil ity of a bullet to acquire striations during the firing event. The SCICLOPS system should be able to measure all bullets no matter the manufacturer. Therefore, the first hypotheses proposed are: H 0 : There will be no differences in the ability of a bullet to acquir e striations based on manufacturer. H 1 : There will be differences in the ability of a bullet to acquire striations based on manufacturer. The second hypotheses deal with the bu llets as grouped by manufacturer. All bullets produced by the same ma nufacturer should be more si milar to each other than to bullets of other manufacturers. This also de als with the SCICLOPS system, because if the bullets of one manufacturer do not match to ot hers of the same type, then the SCICLOPS system will not show the class characteristics that could differentiate bullets of different


29 manufacturers. The system would only show i ndividual characteristics, which is good for matching a bullet to a gun. However, there could be a problem when the gun is not present to make a determination of what kind of gun could have fired the bullet. Therefore, the second hypotheses proposed are: H 0: There will be no differences in the means of measured striations for all bullets of the same manufacturer. H 1 : There will be differences in the means of me asured striations for all bullets of the same manufacturer.


30 Chapter Three Methodology and Data Methodology A secondary data analysis methodology was selected for this project. Secondary data analysis is a research methodology th at involves using data collected by other researchers to answer new research ques tions (Maxfield and Ba bbie, 2001). Although secondary data analysis has several drawback s, including availability, completeness, and validity, this type of methodology is cost-effective and timel y, involving only willingness on the part of the original researcher to allow access to the data. This design allows for further exploration of data already collected, which fits the purpose of this project in assisting in the validation of the SCICLOPS system. Hopefully, this project will allow other researchers to gain access to this valuab le data set and allow for more statistically advanced evaluations of the SCICLOPS system. Data The data for this study comes from the engineering firm of Intelligent Automation, Inc., the creator of the SCICLOPS system. This data was collected for use in testing the SCICLOPS system in the area of gun identifiability. Gun identifiability deals with whether the impressions produced by a guns barrel reproduce the same on every bullet fired by it. The data se t collected by Intelligent Automation Inc. included nine types of bullets in the testing, all of which we re lead core jacketed bullets. A listing and


31 description of all bullet types is given below. Table 2. Descriptions of Bu llets Used in Analysis Manufacturer Caliber Weight Model Magtech 9mm luger 115 Gr. FMC (9A) PMC 9mm luger 115 Gr. FMJ (9A) Remington UMC 9mm luger 115 Gr. Metal case (L9MM3) Winchester 9mm luger 115 Gr. FMJ (Q4172) CCI Blazer 9mm luger 115 Gr. TMJ (3509) Norinco (LY) 9mm luger 124 Gr. China (Ball) Federal American Eagle 9mm luger 124 Gr. Metal Case (AE9DP) Lellier & Bellot 9mm luger 115 Gr. Czech These bullet types were chosen by the rese archers as being of the same type lead core jacketed bullets. The firearm us ed in the analysis was a Ruger P89, whose manufacturing technique was gang broaching. An initial test was completed in which twelve bullets of three different manufact urers (CCI, Remington, and Winchester) were fired by the Ruger into a water tank and retr ieved for analysis, which was a verification that the gun did produce clear and reproducible impressions. Af ter this was verified, ten samples of each type of ammunition were fired. The order of firing was interlaced to prevent bias due to the firing order; t hus, the ammunition was fired following an


32 alternating sequence of types. Therefore, a Magtech bullet was fired first, followed by a PMC, all the way down to LB. Variables Table 3 presents a summary of th e variables used in this study. Bullet1 lists the number assigned to the bullet during the test firings. Magt ech comprises bullets #2 11 (only 10 bullets were tested for Magtech, whic h is one less than all other bullet types), PMC comprises 12 through 21, and so on. Wtavgstr lists the average weighted value for all comparable striations found on the bullet. The average value was weighted in order to compensate for the number of measuremen t pulses taken during the acquisition phase. For whichever reason, during measurement, some of the stria may not have received the same number of measurement pulses from the laser. Weighting the average value allowed for a composite number that took into account the integrity of the measure. This put all of the six measured lands and grooves into equal standing. Relpos indicates the relative position of the bullet on the stage, with 1 be ing the first land or groove measured and 6 being the last. Opticorr indicates the maximum position, which is consid ered the right orientation of the six lands and grooves measured. This variable was used as a comparison point for later significance te sting but has no bearing on the project.


33 Table 3. Summary of Variab le Definitions and Coding Variable Definition Coding Bullet1 ID number assigned to bullet 10 consecutive numbers, starting with 2 and ending with 172 (LB type did not follow exact pattern but consisted of #73 81 and 172 Wtavgstr Weighted mean of the striations measured Any number between 0 and 1, with up to six decimal places Relpos The relative position of the bullet on the analysis stage 1 with 1 being the first land or groove measured and 6 being the last Opticorr The marking of relative positions to show the highest number of striations found between two bullets 0 or 1, with 0 being the incorrect positions and 1 being the highest correlation position (used as a dividing point for significance testing) Analysis Strategy The following analyses were performed on the data set to address the previously stated hypotheses. Descriptive Statistics The first type of analysis presents desc riptive statistics on the weighted average number of striations found on the bullet. Th ese statistics include the number of bullets fired, the number of striations measur ed, the minimum and maximum number of striations found, the average number of stri ations found, and the standard deviation.


34 Normality Testing The second type of analysis presents tables for the assessment of normality in the distributions of each bullet manufacturer. Norm ality testing concerns the examination of each of the distributions (corr ect vs. incorrect orientation) to see whether it violates the assumptions of parametric testing. It is pr oposed that distributions concerning incorrect orientations will be normal or close to normal, if there is nothing operating except random error. The incorrect orientation should not deviate from normality, as the manufacturer type should have no effect on that distribut ion. The correct orientation, on the other hand, is proposed to be leptokurtic, as the weighted number of striations should be highest in this distribution. Manufacturer ty pe may have an effect here, if there are differences in how high the numbers are by ma nufacturer type. This may show that some bullets take striations better than others. This could have an effect on the ability of the SCICLOPS system to measure the striations and make a determination of class and individual characteristics. Significance Testing The third and final type of analysis presents an ANOVA table for each bullet type. As each type contained several bullets, ANOVA was conducted to assess the individuality of the bullet and its manufacturer ; that is, whether the bullets of a certain type showed consistency in the number of striations measured by the SCICLOPS system. This created eight different groups for the eight manufacturers to test for similarity in the means against other manufactur ers and within each manufact urer. Post hoc Tukey tests and homogenous subset tests were performed to examine where any differences existed.


35 Chapter Four Results Descriptive Statistics for Weighted Average Striations by Bullet Type The descriptive statistics for average weighted striations by bullet type are presented in Table 4. Minimum and maximum va lues, as well as the mean and standard deviation, are included. It shoul d be noted that the number of test firings are not equal. The number of striations measured is larger than the number of bullet firings, due to the measurements of 6 orientations of the bullet by the SCICLOPS system. Table 4. Descriptive Statistics for Average Weighted Striations Bullet Type Number of bullets test fired Number of striations measured (N) Minimum Maximum Mean Standard Deviation MAG 730 4470 .240815 .925541 .45244033 .159813317 PMC 629 3870 .246069 .923105 .4469079 .155101147 RUMC 558 3270 .212806 .904019 .43961947 .130965913 WIN 424 2670 .241248 .937230 .45247036 .149304754 CCI 288 2070 .221234 .877064 .41355351 .115519092 NOR 225 1470 .237090 .865893 .44045690 .124976656 FAE 148 870 .288131 .954995 .46271168 .148342660 LB 47 270 .286378 .966533 .44350814 .127912088 All types 3049 18960 .212806 .966533 .44418845 .145197155


36 Normality Tests The distributions of average weighted stri ations showed bi-modality, with a mean of around 0.4 to 0.5, and large amounts of nu mbers on either side. Therefore, the distribution was split in half to show the dist ribution of incorrect orientations, which were proposed to have significantly lower numbers than the corre ct orientation, which would approach 1. A cut-off point of 0.5 was used to separate the distributi ons into incorrect and correct orientations. As each bullet was measured using six orientations, only one orientation, with the highest number of striati ons, was deemed the correct orientation; the other five orientations were deemed incorrect and were therefore expected to have lower numbers than the correct orientation. Analyses therefore concentrated on examining differences among bullet types of the normality or non-normality of their distributions for both correct and incorrect orientations. The incorrect orientation distributions were proposed to approach a normal distribution for every bullet type, due to the low number of striations measured and the occurrence of measurement and random error. The correct orientations would be positively skewed, as they were expected to hover near 1. These distributions were used to show the c onsistency of measurement by the SCICLOPS system. The correct orientation distribution w ould be positively skew ed if the SCICLOPS system was measuring what it intended, to show that all of th e bullets by the same manufacturer were statistically similar in the number of average weighted striations. Examples of these distributions are shown in Figures 1 and 2.


WTAVGSTR.488.463.438.413.387.362.337.312. Dev = .05 Mean = .387N = 15911.00 WTAVGSTR.975.950.925.900.875.850.825.800.775.750.725.700.675.650.625.600.575.550.525.500HistogramFrequency3002001000Std. Dev = .11 Mean = .745N = 3049.00 Figures 1 and 2. Histograms of Incorrect and Correct Orientations for all Bullet Types. As can be seen, the distribution on the left represents the incorrect orientations. A bell-shaped curve can be seen that almost straddles the middle of the graph. The histogram on the right, however, shows a correct orientation distribution. This histogram does not follow a curve, but does look leptokurtic. This is expected due to the higher numbers of striations for the orientation. The analyses conducted were normality tests as well as histograms and Q-Q plots for each manufacturing type of bullet (see Appendix A). The normality test conducted was the KolmogorovSmirnov test with a Lilliefors Significance Correlation. As predicted, most of the bullet type distributions, both correct and incorrect orientations, followed the hypothesized pathway. Tables 5 and 6 show the normality tests by bullet type and orientation. 37


38 Table 5. Normality Tests by Bullet Type for Incorrect Orientation Kolmogorov-Smirnov Bullet Types Statistic Degrees of Freedom Significance MAG .015 3740 .054 PMC .026 3241 .000* RUMC .011 2712 .200 WIN .024 2246 .003* CCI .020 1782 .096 NOR .021 1245 .200 FAE .025 722 .200 LB .047 223 .200 significant at the .01 level The KolmogorovSmirnov test shows that WIN bullets and PMC bullets measured in incorrect orientations do not follow a normal distribution, while the MAG, RUMC, CCI, NOR, FAE, and LB bullets do. Th us, six of the eight manufacturers follow the predicted pattern, while two do not.


39 Table 6. Normality Tests by Bullet Type for Correct Orientations Kolmogorov-Smirnov Bullet Types Statistic Degrees of Freedom Significance MAG .124 730 .000* PMC .090 629 .000* RUMC .054 558 .000* WIN .088 424 .000* CCI .049 288 .098 NOR .107 225 .000* FAE .073 148 .050 LB .150 47 .009* *significant at the .01 level The results for the Kolmogorov-Smirnov test show all but CCI and FAE to be non-normal distributions, with LB near the cut-off point of .01. Thus, five of the manufacturing types followed the predicted pa ttern, while three did not It is interesting to note that the manufacturing types with distributions diffe ring from the expected path for correct orientations were not the same as those differing for the incorrect orientations. Implications of this will be discussed in the next chapter. Analysis of Variance Analysis of Variance tests were performed on each bullet manufacturer, testing across manufacturers as well as within all of the bullets of each manufacturer. The ANOVAs tested differences across the mean s of each manufacturer for the average


40 weighted number of striations. Forty-five cases from each manufacturer were used, at forty-five was the lowest common amount (e qual to the smallest group, LB). This was done to allow for furthe r significance testing. The ANOVAs for both the correct and incorr ect orientations we re significant at the .01 level, F(7, 352) = 50.798, p <.01 and F(7, 352) = 17.620, p<.01, respectively. These showed that significant differen ces existed across manufacturers for both orientations. Post hoc Tukey and Homogenous Subsets Tests Post hoc Tukey tests were then done to examine which manufacturers, if any, had significantly different means from the other manufacturers. Tables 7 and 8 show that there are significant differences across almost all of the manufacturer s at the .01 level. Table 7. Tukey Test Between Manufact urers for Correct Orientations MAG PMC RUMC WIN CCI NOR FAE LB MAG .682 .000* .953 .000* .000* .702 .000* PMC .682 .000* .999 .000* .000* 1.000 .000* RUMC .000* .000* .000* .000* .362 .000* .388 WIN .953 .999 .000* .000* .000* .999 .000* CCI .000* .000* .000* .000* .275 .000* .254 NOR .000* .000* .362 .000* .275 .000* 1.000 FAE .702 1.000 .000* .999 .000* .000* .000* LB .000* .000* .388 .000* .254 1.000 .000* significant at the .01 level


41 Table 8. Tukey Test Between Manufact urers For Incorrect Orientation MAG PMC RUMC WIN CCI NOR FAE LB MAG 1.000 .999 .000* .120 .031 .000* .003* PMC 1.000 .999 .000* .124 .030 .000* .003* RUMC .999 .999 .000* .409 .004* .000* .000* WIN .000* .000* .000* .000* .617 1.000 .951 CCI .120 .124 .409 .000* .000* .000* .000* NOR .031 .030 .004* .617 .000* .418 .998 FAE .000* .000* .000* 1.000 .000* .418 .848 LB .003* .003* .000* .951 .000* .998 .848 *significant at the .01 level As can be seen by these re sults, significant differences exist across manufacturers. No two manufacturers were alik e for both the correct and inco rrect orientations. This was expected for correct orientations, as each manufacturer should have bullets that are not identifiable with other manufacturers. Th e weighted means should be significantly different, to show that the SCICLOPS system does not read every bullet as the same. For the incorrect orientations, how ever, there were some surprising findings. Although all of the manufacturers had one or more other manufac turers to whom they were similar, there were many more significant differences than expected. If the incorrect orientation numbers were due to chance, then there s hould not be significant differences across manufacturers.


The surprising lack of significant differences between manufacturers was also shown when conducting a Homogenous Subsets test, which examines the means for similarity and groups any similar bullet types together. These results are presented in Tables 9 and 10. Table 9. Homogenous Subsets for All Manufacturer Types in Incorrect Orientation mincorr Tukey HSDa 45 .30844982 45 .32125491 45 .32486969 45 .32494656 45 .34438696 45 .34850109 45 .35533316 45 .35710813 .120 .418 Bullet 1 Manufacturer CCLid RUMCid PMCid MAGid NORid LBid WINid FAEid Sig. N 1 2 Subset for alpha = .05 Means for groups in homogeneous subsets are displayed. Uses Harmonic Mean Sample Size = 45.000.a. 42


Table 10. Homogenous Subsets for All Manufacturers in Correct Orientation maxcorr Tukey HSDa 45 .64241969 45 .68938278 .68938278 45 .69022482 .69022482 45 .73327211 45 .85978349 45 .86040164 45 .87165400 45 .89407044 .254 .362 .682 Bullet 1 Manufacturer CCLid NORid LBid RUMCid PMCid FAEid WINid MAGid Sig. N 1 2 3 Subset for alpha = .05 Means for groups in homogeneous subsets are displayed. Uses Harmonic Mean Sample Size = 45.000.a. These findings show further that there is some homogeneity (similarity) amongst the bullet types. For the correct orientation, there seem to be three groupings, with two groups overlapping and then the three group (consisting of PMC, FAE, WIN, and MAG) having higher and non-overlapping means. For the incorrect orientation, there are two groups that do not overlap. None of the subset groups for either orientation are significantly different from others within their group though. These results will be discussed further in the next section. As a whole, these findings support the idea that manufacturers may have an effect on the number of striations measured by the SCICLOPS system. There does seem to be mostly normal distribution for incorrectly oriented bullets no matter the type (with exceptions) as well as non-normal distributions for correctly oriented bullets (again with exceptions). There were significant differences when comparing the average weighted 43


44 mean number of striations for correct orientations across manufacturers, which was to be expected. Differences in Means Within Each Manufacturer Also tested was the hypothesis that ther e would be no differences in the mean average weighted striations for all bullets of the same manufacturer. Only the correct orientations were tested, due to the proposed leptokurtic distributions. If the mean average weighted striations for all bullets of the same type are similar (as they should be), then the ANOVA should not be significant. Table 11 presents these results. Table 11. ANOVA Test for Differences Within Manufacturers Manufacturer N Degrees of Freedom F Significance MAG 746 9, 737 1.381 .193 PMC 646 9, 637 2.683 .005* RUMC 544 9, 535 14.816 .000* WIN 444 9, 435 1.995 .038 CCI 344 9, 335 54.288 .000* NOR 244 9, 235 36.806 .000* FAE 144 9, 135 1.198 .301 LB 44 8, 36 1.761 .118 significant at the .01 level These results show that the PMC, RU MC, CCI, and NOR types have significant differences in the mean weighted average stri ations within each of their bullet types,


45 while the bullets of MAG, WIN, FAE, and LB types do not differ significantly from others within their respective types.


46 Chapter Five Discussion After having examined the results from the data, it is now important to discuss these findings and their implicati ons to the proposed hypotheses. Differences in Striations Measured by Bullet Type The results from the ANOVA show suppor t for both alternate hypotheses, which was that differences exist by bullet manufacturer in the amount of striation measured by the SCICLOPS system and in the differences in means. This can be seen through the ANOVA and even the normality tests. As disc ussed before, the ideal distribution for each bullet type would show a st atistically discernable bim odal pattern, with incorrect orientations having a normal distribution and correct orie ntations having a negatively skewed distribution. The normal distribution of the incorrect orient ations would be the result of random error, while the skewed distributions of the correct orientations would be the result of the clustering of numbers closer to 1. This was not always the case for the eight bullet types tested. When examining the descriptive statistic s for average number of striations, one can see some differences, but the numbers look pretty close t ogether. The highest maximum number of striations was .966533 (LB type), while the lowest was .865893 (Norinco type). The minimum number of striations, mean number, and standard deviations followed the same pattern, with not much visually discernable differences to


47 be found. It is only when examining the histograms and normality tests that the big picture emerges. The histograms of the incorre ct orientations by bullet type showed that visually, all of the bullet types foll owed the predicted pattern. The histograms of the correct orientation by bullet type showed difference s. The Magtech type showed a leptokurtic skewed distribution, as did the PMC type, Win type, and FAE type. Distributions not fitting this pattern included the Remington (RUMC) type, CCI type, Norinco type, and LB type. The LB type had the lowest number of observations to measure (N = 47), which may have affected the tests conducted. Therefore, the LB ty pe will not be discussed in this section. As the histograms were only a visual aid, tests of normality were conducted on both sets of distributions. The results were quite different from what was expected, especially considering the tw o tests performed. The Kolmog orovSmirnov test examines how closely the sample distribution is to normal. When examining normality by bullet type, one can see that the Magtech bullets fo llowed the predicted pa ttern in both cases, with the incorrect orientation distribution ha ving a normal distribution, while the correct orientation did not. This test does not look at why the dist ribution is non-normal (which way it is skewed), just that it does not follow a normal di stribution. The PMC bullet type, on the other hand, had statistically discernabl e departures from normality in both cases. This was unexpected, especially as the histog ram showed no discernable departure from normality. The Remington (RUMC) bullet t ype, showed a normal distribution for incorrect orientations, and a non-normal di stribution for the correct orientations.


48 Winchester (WIN) type showed departures from normality for both distributions. CCI showed normal distributions for both orientat ion types. Norinco (NOR) type showed the expected pattern, with normal distributi on for incorrect orientation and a non-normal distribution for correct orient ation. FAE showed normal dist ributions for both as well. Again, the LB type had much fewer cases th an any other type, and is therefore being omitted from analysis. The pattern that emerged from the resu lts shows that there are differences in bullet type, just not in the pred icted direction. Some bullet type s were very reliable in the amount of striation measured by the SCICLO PS system, while others were not. Some suggested differences would be in manufactur ing techniques and materials used in the construction of that particular type of bullet. Although all of the bullets were of the cast lead jacketed type, manufacturing processes an d type of material used may have differed. The CCI and Norinco types especially s hould be examined for differences in manufacturing techniques. Norinco is a Ch inese ammunition, and perhaps the processes are very different from U.S. manufacturing companies. Th e CCI type showed very low numbers, but this can be due to the fact that the jacketing material is of thicker quality. More information on manufacturin g techniques is needed to conduct further analyses. The ANOVA tests showed further that di fferences exist across manufacturers. In testing all eight of the bu llet manufacturers against each other, the ANOVA showed significant differences, thereby rejecting th e null hypothesis in favor of the alternate hypothesis. In breaking down these differen ces by manufacturer, there seems to be no discernible pattern as to why this is so. Further testing is needed to draw out the


49 differences, for the ANOVA only shows that di fferences do exist, not really why they exist. It may be manufacturi ng techniques, jacket types use d, the order of firing, or some hitherto unknown explanation. This is also seen in the post hoc Tukey and Homogenous subsets tests. Both tests show where exac tly the differences and similarities amongst bullet manufacturing types exist. The Tukey test s show that each of the bullet types is similar to at least one another type, but not to all other types. The Homogenous subsets show that the mean average weighted striat ions of each bullet type are comparable to several others. When comparing the manufactur er types in the incorrect orientation, two distinct groups emerge that do not overla p. Only one group would be expected if the numbers recorded were due to random error. For the correct orientation, three groups emerge when eight separate groups were expe cted. Two of the groups overlap, leading to a conclusion that the CCI, NOR, and LB types are very similar to each other, somewhat similar to the RUMC type, and very different from the MAG, PMC, FAE, and WIN types. The MAG, PMC, FAE, and WIN types all have notably highe r values than the other types. This could also be due to manufacturing processe s or what type of alloy is used to create the bullet jack et. Whichever the case, the CCI NOR, and LB types seem to be read completely differently by the SCIC LOPS system. Further testing is needed to figure out why this is so. Differences Within Manufacturer This is also evident in the within manuf acturer testing. Four of the eight types had significant differences about their means, while the other four did not. PMC, RUMC, CCI, and NOR all had differences within thei r own bullets. These differences show that


50 these bullet types have less identifiability as being of a certain type than do MAG, WIN, FAE, and LB. This could be a problem when trying to match a bullet to a gun by scanning the bullet with the SCICLOPS system. In looking at these results, one comes up with mixed support for the SCICLOPS system. In a perfect world, each bullet manufacturing type would have two distinct distributions comprised of the two orientations and be significantly different from other bullet manufacturing types. Since some of the manufacturing types are not significantly different from each other, it remains a question as to whether class characteristics such as the manufacturing type can be read by the SC ICLOPS system or whether only individual differences can be seen. Finding a syst em that connects already known class characteristics to individual characteris tics should be a major goal in firearm identification. The implications of the ANOV A and Tukey tests are that, although some significant differences could be read by the SCICLOPS system, there were still others that remained similar, and that some of the manufacturer types are identifiable as that type while others are not. More testing in this area should be done with the SCICLOPS system to see if improvements could be made that would increase the number of significant differences found. The goal of identification would be that each manufacturer is significantly different from the rest, making bullets of th at manufacturi ng type easy to distinguish from others. Identifying class characteristics such as that would make identifying individual differences easier, thereby allowing more confident matches between bullets and guns.


51 The question of bullet identifiability has long been of concern in the area of firearms investigation and will continue to be so until a solution is found. Is it that the system cannot measure differences, or does th e problem lie with th e bullets themselves? Future studies should be done with a greater range of bullet types, including not only different manufacturers but also different kinds of bullets (har d nose, jacketed, etc.). This would further test the SCICLOPS system as an accurate identifier of bullets and firearms and allow for a more widespread comparison of how bullets perform against others of the same and different types. Limitations One major limitation of this study is the unequal number of cases for each bullet type. Future testing should comprise equal numbers of bullets, to allow for a more accurate determination of normality and deviation from the mean. Another limitation is the study design. As this is a secondary an alysis, there may be unknown problems with data collection procedures or data qual ity and validity. This may also limit the generalizability of the resu lts and conclusions. The study design included the use of a sequential firing order in which one bullet from each manufacturer was fired in a row before repeating the order. This may have bi ased the resulting numb er of striations found on the bullets. Test firing ten bullets in a row for each manufacturer before going on to the next may produce different results, whic h might prove useful in determining the identifiability of a certain bullet manufacturi ng type. Test firing more than one bullet at a time for a manufacturer may allow striations to appear more uniformly than firing one bullet of each manufacturer at a time, and thus allow for gr eater identifiability.


52 Conclusion In conclusion, this study does find some support for the use of the SCICLOPS system as a technique for firearm identif ication. Several of th e bullet types proved identifiable as a group (of test firings). Uncertainty remain s about whether the lack of identifiability of the other bullet types is due to the identification system or the manufacturing of the bullet. The results suggest that there are two distributions to examine when using the SCICLOPS systemthe incorrect orientations and correct orientations. As these are determined by the orientation having the highe st striation number, there may exist a need to find a more sophisticated technique to ensure accuracy. The SCICLOPS system does in fact represent the next ge neration of firearm identifica tion. This system allows for depth analysis that did not previously exis t with the comparison microscope. Firearms identification is still a large part of forensic s and criminal investigations, and continuous improvements in the SCICLOPS system will show it to be a useful and accurate tool for aiding law enforcement and other firearm experts. Future Areas of Study More research in this area is needed to complete the development of an automated firearms identification system such as SCIC LOPS. Data sets and tests involving various types of firearms, bullets, and even cartri dge cases could strengthen the confidence behind the SCICLOPS system as a useable tool. An area in special need of research is in bullet manufacturing practices. As can be seen in this project, not al l bullets are reliable in their matching to others of the same t ype. Exploration of why this could be is


53 necessary as a backup to further explorations of what SCICLOPS is capable of. If differences between bullet types can be quan tified, then the SCICLOPS system stands a chance at showing identifiability for all type s, no matter the manufacturing process of the bullet. Another area of future study involves a direct comparison of the SCICLOPS system to the comparison microscope and a ny other emerging identification systems. A comparison such as this could prove the im provement of using SCICLOPS over the more traditional method of the comparison microscope. Again, this project concerned only an expl oratory analysis of data created by the SCICLOPS system. More advanced statistical analyses should be conducted to further test the reliability, consistency, accuracy, and validity of the SCICLOPS system.


54 References Bachrach, B. (2002). Development of a 3D-based automated firearms evidence comparison system. Journal of Forensic Science, 47 (6), 1. Biasotti, A. (1959). A statistical study of th e individual characterist ics of fired bullets. Journal of Forensic Science, 4 (1), 34-50. Biasotti, A., & Murdock, J. (1984) Criteria for identification or state of the art of firearm and toolmark identification. AFTE Journal, 16(4), 16-34. Brown, C., & Bryant, W. (1995). Consecutively rifled gun barrels present in most crime labs. AFTE Journal, 27(3), 254-258. Brundage, D. (1994, June). The identification of consecutively rifled gun barrels. Paper presented at AFTE Twenty-Fifth Seminar, Indianapolis, IN. Burrard, G. (1964). The identification of fir earms and forensic ballistics. New York: A.S. Barnes. Churchman, J. (1949). The reproduction of char acteristics in signatu res of Cooey rifles. RCMP Gazette, 11 (5), 133-140. Commonwealth v. Best, 180 Mass. 492, 62 N.E. 748 (1902). Dean V. Commonwealth, 32 Gratt (Va.) 912 (1879). De Kinder, J., Pascal, P., Pirlot, M., & Nys, B. (1998). Surface topology of bullet striations: An innovating technique. AFTE Journal, 30(2), 294-299.


55 Evans v. Commonwealth, 230 Ky. 411, 19 S.W. (2d) 1091, 66 A.L.R. 360 (May 31, 1929). Federal Bureau of I nvestigation. (1981) Handbook of Forensic Science (Rev.ed.). Washington, DC: Federal Bureau of Investigation. Freeman, R. (1978). Consecutive ly rifled polygon barrels. AFTE Journal, 10 (2), 40-42. Galenis v. State, 198 Wis. 313, 223 N.W. 790 (Mar. 5, 1929). Giannelli, P. (1991). Ballistics evidence: Firearms identification. Criminal Law Bulletin, 27(3), 195215. Hall, E. (1983). Bullet markings from consecutively rifled shilen DGA barrels. AFTE Journal, 15(1), 33-47. Hall, J.H. (1900). The missile and the weapon. Buffalo Medical Journal, 39 727. Inbau, F.E. (1999). Firearms identification- Ballistics. Journal of Criminal Law and Criminology, 89 (4), 12931314. Laney v. United States, 54 App. D.C. 56, 194 Fed. 412 (Dec. 3, 1923). Lutz, M. (1970, August). Consecutive revolver barrels. AFTE Newsletter 24-28. Matty, W. (1985). A comparison of three indi vidual barrels produced from one button rifled barrel blank. AFTE Journal, 17 (3), 64-69. Murdock, J. (1981). A general discussion of gun barrel individuality and an empirical assessment of the individuality of cons ecutively button rifl ed .22 caliber rifle barrels. AFTE Journal, 13 (3), 84111. Nichols, R.G. (1997). Firearm and toolmark identification criteria : A review of the literature. Journal of Forensic Science, 42 (3), 466-474.


56 People v. Beitzel, 207 Cal. 73, 276 Pac. 1006 (Apr. 16, 1929). People v. Fiorita, 339 Ill. 78, 170 N.E. 690 (Feb. 21, 1930). Saferstein, R. (1998). Criminalistics: An introduction to forensic science. Upper Saddle River, NJ: Prentice Hall. Skolrood, R. (1975). Comparison of bullets fired from consecutively rifled Cooey .22 caliber barrels. Canadian Sociological and Forensic Sciences Journal, 8 (2), 4952. State v. Smith, 49 Conn. 376 (1881) State v. Boccadoro, 105 N.J.L. 352, 144 Atl. 612 (Feb. 4, 1929). Swanson, C.R., Chamelin, N.C., & Territo, L. (2000). Criminal investigation (6 th ed.). New York: McGrawHill. Williams, Mason. (1980). Practical handgun ballistics Springfield, IL: Charles C. Thomas.


57 Bibliography Agresti, A., & Finlay, B. (1997). Statistical methods for the social sciences. (3 rd ed.). Upper Saddle River, NJ: Prentice Hall. Intelligent Automation, Inc. (2000). Ballistics ma tching using 3D images of bullets and cartridge cases. (Progress Report 4). Rockville, MD: Author. Intelligent Automation, Inc. (2003). Test firing pr otocol #2 for a statistical validation of the individuality of guns using 3D im ages of bullets study. Rockville, MD: Author. Intelligent Automation, Inc. (2003). A statistical validation of the individuality of guns using 3D images of bullets. (Progr ess Report 8). Rockville, MD: Author. Maxfield, M.G., & Babbie, E. (2001). Research methods for criminal justice and criminology (3 rd ed.). Belmont, CA: Wadswort h/ Thompson Learning.


58 Appendix A: Histograms and Boxplots by Orientation and Manufacturer MINCORR.406.394.381.369.356.344.331.319.306. BULLET1= 1Frequency806040200Std. Dev = .03 Mean = .335N = 745.00 MAXCORR.925.875.825.775.725.675.625.575.525.475.425.375HistogramFor BULLET1= 1Frequency120100806040200Std. Dev = .12 Mean = .782N = 745.00 MINCORR.419.406.394.381.369.356.344.331.319.306. BULLET1= 2Frequency706050403020100Std. Dev = .03 Mean = .334N = 645.00 MAXCORR.900.850.800.750.700.650.600.550.500.450.400.350HistogramFor BULLET1= 2Frequency806040200Std. Dev = .12 Mean = .764N = 645.00 MINCORR.450.438.425.413.400.387.375.362.350.337.325.312.300. Dev = .04 Mean = .337N = 545.00 MAXCORR.875.825.775.725.675.625.575.525.475.425.375.325HistogramFor BULLET1= 3Frequency6050403020100Std. Dev = .11 Mean = .696N = 545.00


59 MINCORR.431.419.406.394.381.369.356.344.331.319.306. BULLET1= 4Frequency50403020100Std. Dev = .03 Mean = .346N = 445.00 MAXCORR.900.850.800.750.700.650.600.550.500.450.400.350HistogramFor BULLET1= 4Frequency50403020100Std. Dev = .13 Mean = .749N = 445.00 MINCORR.406.394.381.369.356.344.331.319.306. BULLET1= 5Frequency3020100Std. Dev = .04 Mean = .323N = 345.00 MAXCORR.850.800.750.700.650.600.550.500.450.400.350HistogramFor BULLET1= 5Frequency403020100Std. Dev = .12 Mean = .624N = 345.00 MINCORR.450.438.425.413.400.388.375.363.350.338.325.313.300. BULLET1= 6Frequency403020100Std. Dev = .03 Mean = .345N = 245.00 MAXCORR.875.850.825.800.775.750.725.700.675.650.625.600.575.550.525.500.475.450.425.400HistogramFor BULLET1= 6Frequency3020100Std. Dev = .12 Mean = .678N = 245.00


60 MINCORR.413.400.388.375.363.350.338.325.313.300.288HistogramFor BULLET1= 7Frequency20100Std. Dev = .03 Mean = .352N = 145.00 MAXCORR.950.925.900.875.850.825.800.775.750.725.700.675.650.625.600.575.550.525.500HistogramFor BULLET1= 7Frequency20100Std. Dev = .11 Mean = .767N = 145.00 MINCORR.400.394.388.381.375.369.363.356.350.344.338.331.325.319.313.306.300.294.288HistogramFor BULLET1= 8Frequency121086420Std. Dev = .03 Mean = .349N = 45.00 MAXCORR. BULLET1= 8Frequency1086420Std. Dev = .13 Mean = .69N = 45.00 Boxplots for all manufacturing types 45145245345445545645745N =BULLET187654321MINCORR. 16705156011169513093102258509110111006397271105911209103211017198471092159597183697377717879807759651255298934213475319 45145245345445545645745N =BULLET187654321MAXCORR1. 138311391513837128171359112547130811385513339117011413110225114078839104899925108011024991811110710171984764694759556971177855594151677831673982157465473567157093813781915545549170395143508974114681633177776277666158873067384117352629432143452161218534151711260530431255390134993925325298916577392107426734213847271120134752551