The tools developed demonstrated their effectiveness in four areas. First, they established that the handwriting of twins can be distinguished. Second, an upper bound was provided on the error rate of expert human examiners, since system error rate is lower than that of lay persons but higher than that of experts. Third, frequencies of various letter formation were automatically extracted, so that they can be used in forensic testimony; for example, to state a probability of random correspondence. Fourth, authorship of a historical document was determined. A statistical model for writer verification was developed, so as to allow computation of the likelihood ratio based on a variety of different feature types. Several new algorithms were developed for extended handwriting comparison, including capturing the uniqueness of the writing and presenting results in a nine-point opinion scale that ranges from identified as same to identified as different. A signature verification approach was developed to deal with a small set of learning samples. Extended writing samples such as a paragraph of writing, as well as signatures, were considered. The task of verification, which is to determine whether two writing samples compared side-by-side originate from the same person, was the principal problem addressed. The research led to a U.S. patent, and the software developed has been made available to the QD community by CDs and Internet downloads. 16 figures and a listing of 16 venues in which project results were presented, 39 references