Writer Identification by Professional Document Examiners

Abstract

Reliable data on the capabilities of professional document examiners are scarce, rendering most past characterizations of these capabilities somewhat speculative. We report on a comprehensive test administered to more than 100 professional document examiners, intended to close this data gap in the area of writer identification. Each examiner made 144 pair-wise comparisons of freely-created original handwritten documents. The task was to determine whether or not a "match" was detected, namely whether or not the two documents were written by the same hand. Matching criteria were based on the identification and strong probability definitions of the ASTM standard E1658. The professionals were tested in three groups (in the northeastern, southeastern, and southwestern United States). In addition, we have created a control group of similar educational background. Several individuals training to become professional document examiners were tested as well.Examination of the data and statistical tests show that the answers collected from the professional and nonprofessional groups came from different populations. The trainees' data were shown to have come from a population that is distinct from both professional and nonprofessional groups. Unlike the professional examiners, the nonprofessionals tended to grossly over-associate. They erroneously "matched" many documents that were created by different writers, mismatching almost six times as many unknown documents to database documents as the professionals did (38.3% vs. 6.5% of the documents).The results of our test lay to rest the debate over whether or not professional document examiners possess writer-identification skills absent in the general population. They do.

Forgery Quality and Its Implications for Behavioral Biometric Security

“…We caution the reader that the proficiency of professional document examiners compared to that of laypersons in detection forgeries remains a controversial topic (see for example, [6], [31]). Indeed, while recent studies [11], [12] seem to indicate that a well-trained subset of the population can perform significantly better than chance at this task, these results are still being openly debated. For that reason, we primarily use our analysis of human judges as yet another indication of the importance of using strong forgers for evaluation purposes.…”

An Insight into Forensic Document Examiner Expertise for Discriminating Between Forged and Disguised Signatures

“…It was found that the control group scored a mean of 4.0 (SEM = 1.8) and thus showed a weak ability to differentiate between stimuli, whilst the FDEs scored a mean of 8.5 (SEM = 1.2) and performed significantly better on this task than the control group (independent samples t ‐test, t = 2.465, df = 18, p = 0.024). Thus, there is evidence that subjects can discriminate between forged and disguised signatures, and that the expertise effect of trained FDEs reported in previous studies (1–6) is also evident in this type of task. Of interest, one FDE subject was able to achieve a very high score of 13 on this test (13 correct calls, three inconclusive calls, and zero errors), indicating that with the test signature under study, it was possible for the visual and cognitive system of a highly skilled FDE to very accurately discriminate between the forged and disguised signatures.…”

On the Discriminability of the Handwriting of Twins

“…As professional QDEs have been shown to perform better than lay persons (22), we can also conclude that their performance would be superior to that of the current system. One caveat in our results comparing the performance of humans and machines is that the present testing was on a small set of 12 individuals; a larger scale testing would be needed to compare the absolute performances of QDEs, lay persons, and the machine.…”