Test items, all referencing the same instructional objective, are not equally difficult. This investigation attempts to identify some of the determinants of item difficulty within the context of a first course in educational statistics. Computer generated variations of items were used to provide the data. The results were used to investigate the relationship between item difficulty and the following characteristics: item format, language, calculation requirements, areas under theoretical distributions, and other miscellaneous differences. In general, changes in item content had more effect than changes in format or linguistic structure, and more complex questions were more difficult than simpler questions. It was concluded that item difficulties can be made to fluctuate by changes in how questions are asked, but that further study is clearly needed. (Author/CTM)