An experiment investigated the effects of the instruction rubric at the beginning of an examination paper and of the format of the examination questions themselves. A history test whose questions originally differed widely in format was modified so that its rubric was as clear and concise as possible and all the short answer questions were of the same format. The design of the experiment produced four versions combining the new and old rubrics with the new and old questions so that the effects of the different changes would not be confused. Statistical analysis of data from 200 subjects seemed to indicate that the original version produced the highest candidate achievement scores; however, closer study revealed that the new rubric was in fact leading to better achievement results than the original one. Although the original rubric was clearly superior in the familiar situation, when the situation was less familiar, the new rubric began to show its worth. Readability measures also showed that the new rubric was simpler and easier to understand. Analysis of candidates' choice of questions revealed that question-style may outweigh the actual subject content in the candidate's process of choice and that questions appearing earlier in an examination are more likely to be chosen. These findings indicate that exam writers should consider the effect of the way a question is couched and the importance of readability of instructions. (JL)