Premium
The Reliability of Pattern Classification in Bloodstain Pattern Analysis — PART 2: Bloodstain Patterns on Fabric Surfaces ,
Author(s) -
Taylor Michael C.,
Laber Terry L.,
Kish Paul E.,
Owens Glynn,
Osborne Nikola K. P.
Publication year - 2016
Publication title -
journal of forensic sciences
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.715
H-Index - 96
eISSN - 1556-4029
pISSN - 0022-1198
DOI - 10.1111/1556-4029.13191
Subject(s) - reliability (semiconductor) , artificial intelligence , pattern recognition (psychology) , statistics , psychology , chemistry , computer science , mathematics , physics , power (physics) , quantum mechanics
This study was designed to produce the first baseline measure of the reliability of bloodstain pattern classifications on fabric surfaces. Experienced bloodstain pattern analysts classified bloodstain patterns on pairs of trousers that represented three fabric substrates. Patterns also varied in type (impact, cast‐off, expiration, satellite stains from dripped blood, and transfer) and extent. In addition, case summaries that accompanied each pattern contained contextual cues that either supported the correct answer (i.e., positive bias), were misleading toward an incorrect answer (i.e., negative bias), or contained no directional information (i.e., neutral ). Overall, 23% percent of the resulting classifications were erroneous. The majority (51%) of errors resulted from analysts misclassifying satellite stains from dripped blood. Relative to the neutral information, the positive‐bias information increased correct classifications and decreased erroneous classifications, and the negative‐bias information decreased correct classifications and increased erroneous classifications. The implications of these findings for BPA are discussed.