Summary of Murdock et al.

Introduction

There are two communities, statistics and law enforcement, which have varying opinions of the evaluation of firearm toolmarks. Statisticians are concerned with the absence of probability models that include type one and two error. Through the research that has been done the error rate is indicated to be small. As stated in the articles abstract there are two purposes to the paper: 1.Is to call out the people who claim that their opinions are fact without logic or proper research, 2. Discuss the probability models that exist.

Review of the Current Literature in Statistical Applications for Firearms and Toolmarks

1932 Gunther and Gunther Discussed the need for competent researchers to determine probabilities.

1935 Hatcher and Gunther Explored the development of RMP’s, random match probability (type 1 error). Hatcher described this as a simple task.

1949 O’Hara and Osterburg Evaluated tool marks and gave them a rating.

1949 Churchman Calculated probabilities but didn’t provide information about how he arrived at his conclusions.

1959 Biasotti Conducted a study between known match firearms and known non-match firearms with CMS, consecutively matching striations.

2007 Neel and Wells Known match and known non-match bullets were examined with 2D and 3D technology. They determined likelihood ratios for different CMS runs.

2008 Howitt, Tulleners et al. Explained correspondence probabilities. They used magnification to analyze the CMS’s and were able to obtain a theoretical RMP.

2003 Stone Theoretical RMP rate calculated.

2005 Collins Empirical study confirming Stones theoretical error rate.

2010 Bachrach, Jain, Jung, and Koons Used Confocal magnification to observe striation marks from screwdrivers and pliers. These microscopes would evaluate the marks digitally.

2010 Chumbly et al. Man vs. Machine. A developed algorithm was tested against blind examiners. The examiners prevailed. The algorithm didn’t correctly identify the bullets.

2010 Chu et al. Discusses “automated bullet signature identification using correlation and confocal microscopy”.

2010 Chu et al. Identifies a relationship between striation density and the identification rate.

2011 Chu et al. Discusses the importance of striation edge detection.

2013 Chu et al. Conducted a study that explored a quantitative consecutive matching striae model and a cross-correlation statistical algorithm. 3D confocal microscopy was used. The study was comprised of known samples and blind samples.

2012 Weller et al. Used confocal microscopy to evaluate 90 cartridge cases fired from 10 pistols. He used his data from his study to calculate an error rate. His findings indicate a normal distribution and a very small error rate. When presenting his study at the May 2011 California Association of Criminalists he emphasized that his error rate only worked for the firearms that were used in his study.

2012 Petraco et al. Applied robust identification models to screwdriver striations with barcodes. The error rate remained low.

2011 Gambino et al. Evaluated 58 primer shear marks on 9mm cartridge cases fired from 4 Glock model 19 pistols. Initial findings revealed a low error rate. The study was performed with a confocal microscope.

2013 Petraco, Chan, Phelps, Gambino, McLaughlin et al.
This was a follow up study that expanded on the 2011 Gambino research. The data was expanded to 290 3D screwdriver striations from 29 screwdrivers and 162 primer marks produced by 24 Glocks. Error rates were 0.0001 from screwdrivers and 0.0003 from the Glocks.

2016 Hamby, Norris, and Petraco Conducted an analysis of 1632 9mm Glock cartridges gathered over a 21 year time span. The error rate was 0.0001%. This was done with manual comparisons.

Current State of the Use of RMPs in Firearm and Toolmark Identification

“There is no unifying model for the comparison of striated and impressed tool marks…” There are varying techniques to evaluate the error rate. With each technique the rate remained small. The National Academy of Science states that without quantified random match probability data all expert claims are invalid. Michael Risinger, a member of academia refuted this claim. He reported that error rate modeling for tool marks is difficult because it isn’t comprised of binary units such as DNA. There is also concern in the deterioration of the components of the firearm and the material discharged. Or if the bullets may be treated, e.g. sandpaper, and therefore unrecognizable as being fired from known weapon.

Absolute Versus Practical Identification and Subjectivity

The term absolute certainty is obsolete. Reasonable certainty is the compliment to reasonable doubt. There are many non-subjective measurements that accompany tool marks. Land size, direction of spin, groove counts, and caliber are all non-subjective measurements that contribute to identification of a firearm. Visual observations, such as comparing surface characteristics, are objective and are not measured quantitatively. Thus, in critical examination, visual characteristics are viewed as illegitimate.

Title & Abstract Discussion

This paper discusses the different studies and approaches forensic scientists and statisticians have utilized when examining tool marks produced by firearms. The title “The Development and Application of Random Match Probabilities to Firearm and Toolmark Identification”, is a very direct title. I know what I will be reading. Different approaches to solving the issues contained in firearm toolmark identification. The question that I have when I am reading this title is immediately answered by the abstract. Who are the different parties involved in this discussion? I think that success of firearm tool mark identification lies with both statistics and forensics. Communication between the sciences is vital.