Date of Graduation
Eberly College of Arts and Sciences
Forensic and Investigative Science
When analyzing footwear impression evidence, one of the goals of an examiner is to determine if an exemplar shoe could be the source of an impression found at a crime scene. This opinion is based on an assessment of the similarity of class characteristics and randomly acquired characteristics (RACs) between the known and questioned impressions, as well as the rarity of the observed characteristics. The primary aim of this research was to estimate the random match frequency of randomly acquired characteristics (RAC-RMF) within a forensic footwear database to determine the frequency of RACs with geometric similarity occurring in the same relative position on unrelated outsoles, which could potentially increase the chance of an erroneous source association.
RAC-RMF was estimated using high-quality test impressions of the 1,300 shoes in the West Virginia University (WVU) footwear database. Each impression in the database was sequentially held out and compared to the remaining 1,299 impressions to determine if unrelated shoes possessed similar RACs in the same relative locations. With over 80,000 RACs available for analysis, this resulted in nearly four million comparisons, which were performed using a combination of visual comparisons and predictions from a mathematical model based on a percent area overlap similarity score. Nearly 70% of the shoes in the database shared an indistinguishable pair with at least 1 out of 1,299 unrelated shoes, with a maximum RAC-RMF value of 49 out of 1,299 observed, and up to 5 indistinguishable RAC pairs shared between unrelated outsoles.
A similar evaluation was performed on two simulated crime scene impression datasets each containing more than 160 impressions deposited in blood or dust, respectively. A total of 759 RACs were identified in blood impressions created on tile, leading to over 77,000 non-mated RAC comparisons between blood impressions and test impressions from 1,299 unrelated outsoles. RACs in blood impressions were smaller on average than their test impression mates, and therefore exhibited a 66% increase in the number of indistinguishable RAC pairs. Depending on RAC length, relative RAC-RMFs of at least 0.0008 were encountered at a rate between 3.4% and 34%. The dust impression dataset included impressions deposited on paper and tile, with the latter lifted using either gelatin or Mylar film and an electrostatic lifter. A total of 1,513 RACs were identified from all impressions, generating over 154,000 non-mated RAC comparisons. The RACs in dust impressions were often similar in size or larger than their known mates, leading to a 42% decrease in indistinguishable RAC pairs relative to mated test impressions. As a result, relative RAC-RMFs of at least 0.0008 were observed at a rate between 3.1% and 32%, despite twice the number of RACs available compared to the blood impressions. This contrast suggested that a liquid medium may erode RAC size, while a particulate medium maintains or possibly increases RAC size, thus influencing non-mated RAC similarity. However, no more than one shared indistinguishable RAC pair was observed between unrelated outsoles for either dataset, meaning that an average of four and eight distinguishable RACs were present for blood and dust impressions, respectively.
This research provided estimates of RAC-RMF for a large database of high-quality test impressions as well as two datasets of simulated crime scene impressions. Analysis of these datasets demonstrated that RAC geometries do repeat on non-mated outsoles, and the rate at which this occurs within each dataset was quantified. Since theoretical models have traditionally been the basis for estimating RAC-RMF in footwear and the majority of empirical studies have reported RAC-RMFs at or near zero, the contribution of this research to the forensic footwear community is a calibration of this estimate based on empirical data from a larger sample of outsoles.
Smale, Alyssa N., "Estimate of the Random Match Frequency of Acquired Characteristics in a Forensic Footwear Database" (2023). Graduate Theses, Dissertations, and Problem Reports. 12122.