The 2016 PCAST (President's Council of Advisors on Science and Technology) Report brought forensics, and specifically firearms identification, into the spotlight once again. According to the PCAST Report, the field lacks empirical data supporting firearms examiners' ability to determine what source (i.e., firearm) created the markings left on fired bullets and cartridge cases.
In response to this criticism, Houston Forensic Science Center (HFSC) designed a black box study using ten consecutively rifled and three nonconsecutively rifled Ruger LCP barrels. Like many consecutively rifled barrel studies, test kits consisting of known samples (i.e., test fires) and unknown bullet samples were assembled and distributed to a broad group of firearms examiners across the Unites States and internationally. In this study, four different open-set test kits were created. Participants were asked to choose one of five possible conclusions for each comparison: identification, inconclusive A (leaning toward ID), inconclusive B (neutral), inconclusive C (leaning toward elimination), or elimination. The examiners' results were used to establish (an) error rate(s) for this study.
Unlike other consecutively rifled barrel studies, HFSC incorporated 3D technology to add a second prong to this study. HFSC teamed up with Dr. Alicia Carriquiry and Dr. Heike Hofmann of the Center for Statistics and Applications in Forensic Evidence (CSAFE). A sample of each of the four test kits was sent to Iowa State University, where the CSAFE team scanned each known and unknown bullet using a confocal light microscope. Based on these 3D scans, CSAFE researchers used a RandomForest score to establish similarity scores between the land engraved areas (LEAs) and predict source relationship between bullets. An error rate was established for this method of identifying a common source. CSAFE's similarity scores were compared to the examiners' results. The scoring of inconclusive results was of particular interest: there was no clear separation between inconclusive and elimination scores. However, there was a distinct separation of identification scores, as expected, highlighting the potential for the use of 3D technology in support of identifications in casework.