Twenty multiple-choice vocabulary items and 20 free-response vocabulary items were administered to 660 college students. The free-response items consisted of the stem words of the multiple-choice items. Testees were asked to respond to the free-response items with synonyms. A computer algorithm was developed to transform the numerous free-responses entered by the testees into a manageable number of categories. The multiple-choice and the free-response items were then calibrated according to Bock's polychotomous logistic model. The study demonstrated that vocabulary items presented in a free-response format can provide more information than similar items presented in a multiple-choice format. Two probable sources of this superiority are: (1) obtaining a correct response by guessing is not possible using the free-response format, and information lost, due to the uncertainty about whether testees answered correctly because they knew the answer or because they guessed, is recovered; and (2) more latitude in degree of correctness is present in free-response items than is typically present in multiple-choice items and a testee's degree of partial knowledge is easier to assess. These advantages apparently overshadowed any deficiencies present in the machine-scoring algorithm employed. (Author/MV)