There have been many attempts to formulate a procedure for extracting information from incorrect responses to multiple choice items, i.e., the assessment of partial knowledge. The results of these attempts can be described as inconsistent at best. It is hypothesized that these inconsistencies arise from three methodological problems: the difficulty of the test, the use of omits in scoring, and the fit of the items to the scoring model. This paper reviews the impact of these three problems on the assessment of partial knowledge and proposes a model which will lessen their impact on the evaluation of any partial knowledge scoring model. (Author)