Research

Imagine: you receive a threatening letter and you want to find out who wrote it. Or the police finds a note at a suicide scene, and there is doubt whether or not it was really produced by the suicider. There are different things one can focus on to find the identity of the writer: finger prints, traces of DNA, language usage, but also handwriting. Whose handwriting is on the paper?

Practice

Handwriting research is often done by forensic institutes. In most of these institutes computers are used to assist the human handwriting experts, but often this involves only keeping track of databases. Smart computers that can help humans in the actual handwriting comparison are still hardly ever used. The research I perform to get my PhD-degree is about using computers in forensic handwriting analysis.

Identification

If you have absolutely no clue about who wrote a certain letter, writer identification can be used. You can search for similar handwritings in a large database of documents of who you know who produced them. This can only be done by comparing the unknown handwriting to each of the documents in the database. If this database is really large (let's say that it contains 100,000 documents), this task cannot be performed by human experts in reasonable time. If the computer would be able to make a pre-selection of for example one hundred documents, the task of human writer identification becomes feasible.

Verification

But computers can also be of use when a one-to-one comparison needs to be done, for example in the suicide scenario that was mentioned above. Imagine that the forensic expert finds a certain character shape in both documents (in the suicide note and for example a shopping list of which we know that the possible suicider wrote it). In this situation it is important to know how special that shape is. If 10 percent of the population uses that shape, it is not so distinctive and tells us little about the chance that the two documents were produced by the same writer, while if only 0.01 percent of the population uses the shape, the situation is entirely different! And this effect is even bigger when you concentrate on combinations of shapes that occur in handwriting.

Using a computer, the expert can quickly analyze the handwriting of large groups of people, and get statistical information about how special a character shape or a combination of character shapes is. Given this information, it is easier to gain a more objective and robust conclusion about the chances that the two documents were, or were not, produced by the same writer.

My research

I perform my research at the Donders Institute for Brain, Cognition and Behaviour (part of the Radboud University Nijmegen), where I concentrate on the automatic analysis of character shapes and combinations of character shapes. This means that my techniques do not take into account things like the distance between words and lines, the placement of commas, full stops and hyphens, or the skew of the writing (within my project, these aspects are taken into account by my colleagues at the University of Groningen and the Netherlands Forensic Institute). I have developed a technique that is able to do character comparison in a human like manner, so hopefully the forensic computer systems of the future will correspond to the expectations of human handwriting experts and will seamlessly integrate with their daily practice.

Graphology versus graphonomy

Conclusions that forensic experts draw, and on which my techniques anticipate, are only about the chance that two documents were produced by the same person. These are conclusions. This is completely different from what graphologists do: they try to match personality traits to handwriting (e.g. "this person puts the bar through the 't' relatively low, therefore he probably has a low self esteem"). Graphonomists believe that this direct connection between handwriting and personality does not exist, and since there is no scientific evidence for it, I stay away from it.

More information

Of course, four years of research cannot really be summarized in only a couple of paragraphs. For that reason I am working on a doctoral thesis, in which my research is described in detail. If you are interested in my research, please take a look at my publication list on this website. Here you will find the articles on which my thesis will be based.