4-option questions, but no such overall tendency for 5-option questions.3 Forty-five percent of the sources were biased in their use of this option, but they were skewed in both directions, thus tending to offset one another in the summary study-one-third strongly favored that choice and two-thirds pointedly avoided it. There was a strong bias towards the "all of the above" choice in both 4-option and 5-option questions: 69 percent correct for the former and 54 percent correct for the latter, where the expected values are 25 and 20 percent, respectively (see Table 1). Eighty percent, of the individual sources favored this as the correct option. For true-false questions, since a "true" statement may appear obviously true to the author, he may have a bias towards writing "false" statements. Moreover, due to the structure of the question, the statement must be completely true to qualify as "true," whereas if it is partly false, the correct choice is "false." Thus, one might expect a bias towards "false" answers. The overall data on true-false frequencies rejected our initial hypothesis. Instead, there was a mild bias towards the "true" option.4 The aggregate data, however, concealed considerable bias among individual authors. Twelve of the sources (41 percent) indicated significant bias: 7 favored the "true" and 5 the "false" option. Some authors were very sharply skewed from the norm on true-false questions. Since an examination is a sample of a student's knowledge, the validity of that sample is diminished if it contains the type of bias described above. Users may correct this deficiency through careful selection, rearrangement, or rewriting, while a careless user may increase the bias.