Two basic issues are discussed that involve the Carver model of the relationship between test score and knowledge. The first of these concerns different types of tests which have different probabilities of obtaining the correct answer due to guessing. A derivation is shown which places different test types "on-a-par"--meaning that two tests are designed to maintain a comparable spread in scores; more specifically, this is defined such that two tests have approximately the same rate of change in test score as the proportion of knowledge changes. An equation is provided to achieve this which compares a true-false test to multiple choice or matching tests. It is concluded that for two tests to be "on-a-par" the number of questions must differ, given the probabilities of obtaining correct answers by guessing differ. The second issue discussed involves approximate confidence intervals on the proportion of knowledge obtained. An equation is provided which gives an approximation for the standard deviation of the proportion of knowledge; limits for three standard deviations appear in another equation. This connection of test scores and knowledge level is very important. It provides a standard basis for evaluating performance. In addition, it provides some insight into teaching effectiveness. (RC)