Measurement error in assessment

Multiple-choice tests are usually scored dichotomously, i.e. as correct or as incorrect. A correct response is scored one and an incorrect response is scored zero. Using these scores different psychometric paradigms and models are used to analyse the data and to quantify the performance of the test takers. Classical Test Theory commonly add the question scores to obtain a total score whilst Rasch and Item Response Theory models estimate measures from probabilities. In this paper an argument is made that dichotomous scoring includes significant measurement error as uncertainty of responses is not considered. It is demonstrated how Option Probability Theory can overcome this through assigning percentages to one or more options according to the test taker’s mental processes.