Controlling the quality of multiple choice tests

In Italian Universities, multiple choice tests, an examination methodology particularly popular in the United States, is becoming not only an admission or vocational guidance tool, but also a form of examination within the course itself. The University of Verona has for long been developing and utilizing complex methodologies, like item randomization and optical scanning providing automatic score conversion to process multiple choice tests . Aim of this article is to verify similar performances on the same Business Management multiple choice test, taken by three groups of students from different faculties: law, training science, and computer science. The comparison methodology is based on inductive statistics (Chi Squared tests and One-Way Anova). This work has been one of our efforts to control the quality of the assessment services offered to the students by the Docimological Center. No significant performance differences between the groups have been noticed: it can mean either that the students coming from the three different faculties can be considered equivalent regarding aspects of learning such as intelligence and motivation, or that the employed performance assessment test is not able to evidence any performance difference. We are aware of the current early stage status of this research, so we plan to continue it in order to keep making sure that the multiple choice tests being developed are eligible to be considered reliable ones. The long term outcomes likely to be pursued are specific reliability measurements and improved test quality. The number of relevant performance-related variables is potentially huge. However, that has not discouraged us and has lead us to collect further variables (i.e. personal data) on dedicated questionnaires, which will probably enlighten us for future results that will be preferably obtained by mean of a multivariate statistics approach.