In order to improve understanding of the individuality of handwriting, this study compared handwriting samples of twins and nontwins, using an automatic handwriting verification system.
Results of the verification tests showed that the handwriting of twin siblings was less distinctive from one another than was the handwriting of nontwin siblings. The verification error rate for twins was 12.91 percent compared to 3.7 percent for nontwins. Within the twin sample, error rates for identical twins were higher than for fraternal twins. Error rates in all cases could have been reduced by having the option of rejecting (not making a decision on) borderline cases. The automatic handwriting verification system yielded error rates lower than that of laypersons but higher than that of questioned document examiners based on previous relevant research. With further system improvements, the automatic verification system may achieve the performance of qualified document examiners. Handwriting samples were obtained from 206 twin pairs. Each twin provided a page of handwriting. The verification task consisted of determining whether two half-page documents (original samples divided into upper and lower halves) were written by the same individual. For twins, there were 1,236 verification cases, including 824 tests in which the text content of the writing was different and 412 tests in which the textual content was the same. An additional 1,645 test cases were obtained from handwriting samples of nontwins. In order to make the handwriting comparisons, the automatic handwriting verification system computed overall pictorial attributes (macro features), characteristics of individual letters (micro features), and characteristics of whole-word shapes and letter pairs (style features). Four testing scenarios were assessed: twins and nontwins writing the same text and writing different texts. 14 tables, 16 figures, and 22 references