Regional and National Variation in Arabic Handwriting

Abstract

It has been established in a number of research publications that a
careful study of general handwriting features based on class characteristics
could indicate either the place or the country where the writer was first taught to
write. Using these studies as the basis, this research was carried out explicitly
to understand the characteristics of Arabic handwriting. The aim of this study
was to determine the presence of any particular features or characteristics that
may be common to individuals of a given region or nationality. This was done by
obtaining samples of handwriting collected from individuals of four countries
including; Jordan, Morocco, Oman and Tunisia, where Jordon and Oman are
considered to be Eastern Arab world and Morocco and Tunisia in the Western
Arab world.

An attempt was made to establish whether it was possible to determine
either the region or nationality of the writer of an Arabic passage of text, based
on the formation and the style of the handwriting using specific Arabic
characters. Different steps were taken towards the identification of the class
characteristics of Arabic handwriting in this study starting with the collection of
600 handwriting samples from the participants in four Arabic countries
employing; 150 handwriting samples from each. Ten different characters and
one word were selected for examination, with more than one form of each
character in different positions being identified and the handwriting samples
classified accordingly. In total, 221 class characteristics were identified from the
samples based on different criteria including the shape, number of strokes, pen
movement and starting point.

Tests of association using chi-squared on individual characters showed
that the p-value is less than 0.001 in every case. Correspondence analysis was
used to produce a plot of relative similarities where the different countries
appear as discernible, but overlapping groups. ANOSIM showed these groups
to be statistically different (R = 0.321 p = 0.0002, 1000 permutations). Tree
analysis was used to create a classification system and blind tests were
conducted to test the accuracy of the classification system. On the basis of the
statistics used, significant differences were found in character forms used by the
individuals from the four Arabic countries, suggesting that either region or
nationality of the writer may potentially be predicted with a useful degree of
accuracy. Though the samples were obtained from only four countries out of a
total of 22 Arab countries and only ten characters and one word out of 28
characters were chosen in this study, the results obtained are valuable and
useful, particularly to Forensic Document Examiners (FDEs). In turn this could
be implemented in practice in a situation where a questioned document
containing Arabic text is presented and the suspected author could have come
from one of the four considered countries.