In the past, several latent structure models have been proposed for handling problems associated with measuring the achievement of examinees. Typically, however, these models describe a specific examinee in terms of an item domain or they describe a few items in terms of a population of examinees. In this paper, a model is proposed which allows a population of examinees to be described in terms of an item domain. As is illustrated, the model yields a new formula score which allows guessing to vary over the population of examinees.