A novel scoring procedure was investigated in order to obtain scores from a conventional multiple-choice test that would be free of the guessing component or contain a known guessing component even though examinees were permitted to guess at will. Scores computed with the experimental procedure are based not only on the number of items answered correctly, but also on the average quality of both correct and incorrect choices as reflected in the difficulty and discrimination values associated with these choices. The scores resulting from this procedure were compared to number-right and conventional formula scores for predicting guessing-free scores, which are independently determined for the same test. These data suggest that significant increases in reliability can result if score credit is assigned only to items that were answered correctly without guessing on any type. Since the examinees were not aware that the statistical analysis of their tests included a penalty for guessing, findings should be treated with theoretical interest. (Author/BJG)