The complexity of a handwritten signature is a key factor in the ability of forensic document examiners (FDEs) to determine whether a questioned signature is genuine or a simulation. Examiners are likely to offer stronger and more reliable conclusions in comparisons involving high complexity signatures. Comparisons of low complexity signatures introduce uncertainty and are likely to result in weaker conclusions or perhaps an opinion that the evidence is inconclusive regarding authorship. It is thus of great interest to explore the reliability and reproducibility of perceptions of signature complexity. This article reports on a study which collected signatures from 123 individuals and obtained subjective assessments of the complexity of the signatures on a three-point scale and a five-point scale from five FDEs. The degree of correspondence of the complexity ratings was evaluated by considering each pair of examiners separately. There was exact agreement on the five-point scale for about 45% of signatures and the examiners differed by more than one point on the five-point scale for approximately 9% of signatures. A statistical model treating the scores as continuous measurements estimates the intraclass correlation, the expected correlation of complexity scores from two different FDEs, at 0.65. Data on intra-rater reproducibility of complexity assessments was obtained by acquiring additional evaluations of complexity on seven of the signatures from the same FDEs. These data suggest the expected intra-examiner correlation of complexity scores aross the two occasions is 0.67, slightly higher than the inter-examiner correlation. These results regarding repeatability and reliability of subjective assessments of complexity provide important information for researchers developing objective measures of complexity to enhance the signature examination process.
Assessing the complexity of handwritten signatures
Journal: Law, Probability and Risk
Primary Author: Hal S. Stern
Secondary Authors: Miriam Angel, Melvin Cavanaugh, Eric Lai, Shuying Zhu