A statistical study of the relative difficulty of freehand simulation of form, proportion, and line quality in Arabic signatures Abdulaziz Al-Musa Alkahtani, Andrew W.G. Platt ⁎ Department of Forensic Science, Faculty of Science, Staffordshire University, College Road, Stoke-on-Trent, ST4 2DE, UK abstract article info Article history: Received 27 April 2009 Received in revised form 13 August 2009 Accepted 24 August 2009 Available online xxxx Keywords: Arabic signatures Simulation Statistical analysis This article compares the skill with which a large sample of Arabic writers was able to simulate different handwriting elements in two Arabic signatures. The data agree with the consensus of experts about Roman script: Form is significantly better simulated than Proportion, and Proportion is significantly better simulated than Line Quality. The fact that patterns long observed in Roman script simulation can be statistically demonstrated to exist in the simulation of Arabic signatures suggests that these may be widespread, if not universal, patterns of simulation among writing systems. This is also supported by similar results from a study of abstract characters (Leung et al. 1993), which suggests that these tendencies may even transcend linguistic writing systems. Among the Proportional elements are slant, alignment, spacing, and size. © 2009 Forensic Science Society. Published by Elsevier Ireland Ltd. All rights reserved. 1. Introduction Consensus exists among forensic document examiners (FDEs) about the relative ease with which most handwriting elements can be simulated, but these assumptions had not been validated in controlled experiments. In recent decades, questions have been raised about the reliability of the assumptions underlying forensic handwriting analysis, owing to the lack of scientific testing of much of this knowledge. In 1993, the United States Supreme Court ruled that scientific evidence admitted in trials must be backed by scientific testing of the theories on which the techniques are based, error rates of the techniques, peer reviews of the tests, and acceptability in the relevant scientific community (Daubert v. Merrell Dow Pharmaceuticals 1993). Since that finding, U.S. courts have sometimes refused to admit handwriting analysis and other forensic methods, including fingerprint comparison, into evidence. These deci- sions are probably unfortunate from the viewpoint of justice, as critical evidence, based on techniques informally tested in long practice, may have been denied to jurors. But the current mood of skepticism offers an opportunity to forensic document examiners (FDEs) to test their assumptions and methods with controlled, statistical studies, and in the process to shed light on details on which there may have been disagreements. Thus, experimental examination of the assumptions and techniques of forensic document examination is an important and growing area of forensic document examination, to which the present research contributes. This study applies statistical methods of comparison and probabil- ity estimation to one of the basic questions of forensic handwriting analysis: Which elements of handwriting are easiest and most difficult to simulate? The method used in this study was to elicit freehand simulation of Arabic signatures, under controlled conditions, by a large sample of Arabic speakers and writers, and then to compare the quality of the simulation. The phenomenon of freehand simulation is addressed in a majority of the texts and articles in forensic document examination, mainly because most forgeries are of this type. Another reason for this choice is that freehand simulation tends to produce errors in a great many elements, in contrast to traced simulation, which is detected chiefly by its characteristic line-quality errors. The simulation errors in a wide variety of elements are, of course, desirable in this experiment. The study of signatures is also the focus of most studies of simulation, and central to the field of forensic document examination, because signatures are involved in most forgeries. For that reason, signatures were chosen over other types of text in this experiment. A unique aspect of this study is its focus on simulation of Arabic signatures by Arabic-writing participants. The literature on hand- writing simulation deals almost entirely with simulation of Roman script, in which European languages are written, and predictions drawn from that literature are tested on Arabic script in this study. If the Roman script assumptions are supported by the Arabic script data, then we might hypothesize that the same assumptions are true for other scripts as well. Arabic writing, like much Western writing, is cursive, and both scripts are characterized by a fairly continuous line, broken between words. (There are six letters in Arabic that are also followed by mandatory breaks, even in the middle of words.) Mature, habitual writers of cursive Roman script produce a smooth Line Quality with unconscious rhythmic changes in pen pressure, all of which are lost when they carefully copy the shapes of another person's handwriting style [10,11,13,16,18,22]. It seems likely that Science and Justice xxx (2009) xxx–xxx ⁎ Corresponding author. E-mail addresses: almusa44@yahoo.com (A. Al-Musa Alkahtani), a.platt@staffs.ac.uk (A.W.G. Platt). SCIJUS-00165; No of Pages 5 1355-0306/$ – see front matter © 2009 Forensic Science Society. Published by Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.scijus.2009.08.003 Contents lists available at ScienceDirect Science and Justice journal homepage: www.elsevier.com/locate/scijus ARTICLE IN PRESS Please cite this article as: A. Al-Musa Alkahtani, A.W.G. Platt, A statistical study of the relative difficulty of freehand simulation of form, proportion, and line quality in Arabic signatures, Sci. Justice (2009), doi:10.1016/j.scijus.2009.08.003