Computer Recognition of Handwritten First Names

Alearning program incorporating a version of stimulus-sampling theory was prepared for a digital computer. Handwritten first name signatures coded in a 20 by 48 grid served as inputs to study the effects of parameter changes in the program. Under one main condition the computer with its program examined only those grid cells in which any part of a given pattern fell. Under a second main condition each of the 960 grid cells was examined. Learning of correct pattern names occurred under both conditions. The level of learning achieved under the second condition was higher, although the improvement was gained at the expense of considerably extended running times.