ABSTRACT
In the stage of test construction following the construction of the test specification, we will be generating a large number of possible items. There is then the need to reduce these to a manageable number, and in doing so, to select the best. But how are we to make such a selection? The answer is provided by a procedure known as item analysis. Within the item analysis all the possible test items are subjected to a stringent series of evaluation procedures, individually and within the context of the whole test. This process takes place in two stages: first, a review by experts and second, a pilot study. In the first stage each item should be discussed with an expert, someone who knows and is familiar with the subject matter in question. The rationale for each item should be made explicit, and the stages of reasoning which the respondent may follow should be rehearsed. If the item is to be scored as right or wrong, then possible alternatives need to be explored, as it is occasionally the case that respondents may (correctly) be able to identify the distracters as correct under certain circumstances. The more care that is taken at this early stage the better, as once the test has been piloted we no longer have a completely free hand in making changes to the items, except by carrying out a re-pilot at further expense.
