The suggestion that multiple-choice items can be converted to true-false items without essentially changing what the item measures and with possible improvement in efficiency is investigated. Each of the 90 four-choice items in a natural science test was rewritten into a pair of true-false items, one true, one false. The resulting 180 items were divided to make two 90-item forms A and B which were administered to chance halves of a class of college students. Following item analysis, the most highly discriminating member of each pair of items was chosen for further comparison with the multiple choice forms. Using these selected T-F items, two additional experimental forms, half true-false and half original multiple choice items, were then constructed and administered. Analysis of the resulting data indicates that true-false test items, item for item, are less discriminating than multiple-choice items. This gives partial support to the belief that minute for minute a true-false test can be as reliable as a multiple choice test. It also indicates some support to the hypothesis that there is no important difference in what the two item forms measure. Overall results, despite their limitations, tend to strengthen rather than weaken faith in the usefulness and value of true-false test items. (LR)