LIMA: Forensic Handwriting Analysis
Handwriting analysis has the unusual distinction of being an area of interest in both literature departments and forensic science. Professional forensic document examiners have produced a very substantial body of work, of which I only scratch the surface. Book length introductions to forensic document examination include:
- Wilson R. Harrison, Suspect Documents: Their Scientific Examination, 2nd edition (London, 1966)
- Roy A. Huber and A. M. Headrick, Handwriting Identification: Facts and Fundamentals (Boca Raton, FL, 1999)
- Ron Morris, Forensic Handwriting Identification: Fundamental Concepts and Principles (London, 2000)
One issue that forensic handwriting analysts often confront is the possibility of forgery.
Levels of Proof and the Reliability of Handwriting Analysis
Comparing samples of handwriting does not necessarily give a straightforward unambiguous result. Uncertainties about what may be a style characteristic, the quality of the samples, and the likely degree of variation, means there is often a degree of uncertainty. So how fallible is handwriting analysis?
Handwriting analysis comes under scrutiny when it is used as evidence in court. Tom Davis has written an article on Forensic Handwriting Analysis in Britain, which describes the level of care in accumulating and presenting evidence, and attention to wording in summarising conclusions, which is demanded of the expert witness.
More systematic attention has been paid to the methodological basis of handwriting analysis in the USA, where in 1993 the Supreme Court handed down its decision in Daubert v. Merrell Dow Pharmaceuticals, which set new criteria for the admissibility of scientific evidence, later expanded to include all expert opinion testimony. The Supreme Court formulated a set of factors about proposed testimony that a presiding judge should consider in order to determine "the scientific validity and thus the evidentiary relevance and reliability of the principles that underlie a proposed submission." These factors include:
- Whether the theory of technique can be and has been tested.
- Whether the theory or technique has been subjected to peer review and publication.
- The known or potential rate of error.
- The existence and maintenance of standards controlling the technique's operation.
- Whether the theory or technique is generally accepted within the relevant scientific community.
Evidence must be shown to meet these criteria before it can be presented in court. The ruling placed considerable pressure on handwriting analysis to prove that it was a genuine form of expertise according the Daubert criteria.
Some years previous to the Daubert ruling, D. Michael Risinger, Mark P. Denbeaux, and Michael J. Saks published an article with the striking title, 'Exorcism of Ignorance as a Proxy for Rational Knowledge: the Lessons of Handwriting Identification "Expertise"', University of Pennsylvania Law Review, 137 (1989), 731-92. It accused handwriting analysis of being a pseudo-expertise, its practitioners of being reluctant to allow their work to be tested independently, and of failing to show an acceptable level of accuracy in the few empirical studies that had taken place.
Handwriting analysts have responded to these challenges in a number of ways. There have been further tests on the reliability of analysts' conclusions. An interesting study highlights the problem of false matches: Moshe Kam, Gabriel Fielding, Robert Conn, 'Writer Identification by Professional Document Examiners', Journal of Forensic Sciences, 42 (1997), 778-86. Kam et al. conducted a test on both professionally trained handwriting analysts, and a control group. The study revealed a statistically significant difference in preponderance to make type-I errors (false matches). All groups performed roughly equally in detecting matches, doing so about 88 per cent of the time; however the wrong association rate of non-professionals was about 38 per cent - compared to under 7 per cent among professionals. This difference may well be linked to the methodological difference noted before: professionals start by looking for differences between samples, non-professionals tend to base their conclusions on similarities. We would do well to bear these results in mind when assessing published analyses.
Another interesting recent development, and one that gives support to its objective testability, is the development of computer technologies for handwriting analysis (known as FISH), which are based on the fact that a unique set of algorithms can be generated by performing certain measurements on an individual's handwriting. Work on handwriting individuality has been done by The Center for Excellence in Document Analysis and Recognition (CEDAR), and their findings can be found on their website, where you can even try out a Handwriting Verification Test. CEDAR claim that their computerised analysis can correctly identify an individual's handwriting with 98% accuracy when there is an adequate sample.
There has not been a consistent decision by judges over whether handwriting analysis meets the Daubert criteria. Some judges, such as in a 1999 ruling in Massachusetts (this and other case reports are found on www.forensic-evidence.com), have allowed testimony about (dis)similarity, but not conclusions about authorship. The Mass. judge noted that because an individual's handwriting varies each time he or she writes (unlike, say, a fingerprint), analysis depends on a judgement of similitude that is ultimately subjective. Although an expert's experience makes them better qualified than a lay-person to find similarities, this expertise did not give them any additional qualification to make the next step - identification of authorship. This was therefore left to the jury. The judge did not accept that studies such as Kam's have 'established the validity of the field'.
Other rulings, however, have given greater credence to recent studies of handwriting analysis and seen greater significance in the extensive professional training of expert analyts, and so many judges have accepted that the discipline meets the Daubert criteria. The expertise of those who have attempted to discredit handwriting analysis (eg Risinger, Denbeaux and Saks, none of whom are themselves trained in handwriting analysis) has also come into question. For example see a 1999 case report, and especially the 2002 'Prime', and the similar 2003 'Thornton' cases.
The Relevance of Forensic Analysis to Scholarly Analysis
Since the vast majority of work on handwriting analysis comes from the forensic field, it is clearly useful for anyone dealing with questioned handwriting to have some awareness of forensic work. However there are significant differences between the fields.
For example, forensic document examination has considerably more resources available than does research in the humanities, and few of those who publish on handwriting in the humanities can be considered professional analysts. The levels of rigour found in forensics could not possibly be sustained in the research environment of the humanities.
More important still is the difference in the burden of proof. In the Anglo-American criminal justice system, proof must be established beyond reasonable doubt, but can we really expect a bibliographer, historian, or literary scholar be expected to meet the same criteria of proof? There is a great deal more at stake in a criminal case than in an academic article, so it is surely reasonable to expect more rigorous demands. No-one goes to prison on the basis of a badly argued academic article.
Scholarship in the humanities does not proceed on the basis of establishing its claims to the non-specialist beyond reasonable doubt; it is rather a matter of positing a viable hypothesis to a specialist audience, to whom it will be accepted in the absence of any viable alternative. This demands a lower level of proof. A classic example is the general (but not universal) acceptance of "Hand D" as Shakespeare's. This would not stand up in a law court, but with the support of other (also inconclusive) lines of evidence, and in the absence of a more convincing alternative, it has been sufficient to convince a majority of the scholarly community.
It is reasonable to accept - cautiously - a scholarly identification of handwriting which depends on a balance of probability. However the scrutiny which forensic analysis has undergone should help us to maintain a healthy scepticism about handwriting identification, especially when a document is simply asserted as being in a given person's handwriting without the basis of this identification being made clear.
Handwriting links:
CEDAR's research site on the individuality of handwriting