Item response theory (IRT) is based on the assumption that a direct relationship exists between an examinee's total performance on a set of items and the difficulty of each item on the test. The Rasch model represents this relationship mathematically on an equal interval scale. This paper argues that IRT, under the required conditions, provides measurement comparable to that used for experimentation in the natural sciences. IRT is a unification theory in that it brings two variables together in a mathematical relationship, and it is a probabilistic model because the nature of the data allows prediction of probability of success on any calibrated item, given the student's achievement in Rasch units. The Rasch item calibrating model is tested through demonstrations of experimental methods and results. The most important use of the Rasch model is to produce an extended equal interval scale on which basic skills items are calibrated for an entire elementary school continuum. The main reason for creating a Rasch calibrated item bank is to make possible the construction of different levels of tests, all yielding comparable results on the same scale. Studies of the Portland (Oregon) Levels Tests supported three hypotheses about Rasch model calibration: (1) that item bank calibrations do, with a few exceptions, remain stable across grade level groups performing on the same items; (2) that item calibrations remain consistent from year to year and from grade to grade given appropriate testing and calibrating procedures; and (3) that for the same items, Rasch-calibrated by the fixed parameter model, administered to different groups of students, all calibrations for each item will be the same. These results support the use of Rasch model calibration when proper precautions are taken. (Contains eight figures.) (SLD)