The feasibility of criterion referenced testing is held to be dependent on the tenability of two postulates: (1) that bias can be controlled in a principled manner from one test to the next; and (2) that one mental process measured by such tests may lawfully interact with another. Without the first postulate, criterion scores could not be established rationally. Without the second postulate, the usefulness of reaching the criterion score could not be established. These concepts apply both to content mastery tests and to task proficiency tests. Two corresponding research hypotheses are the hypothesis of regression identity for lawfully made tests and regression regularity. Results of earlier research on cloze tests, reading comprehension completion tests, and three preference scales relating to reading materials administered in grades 3-12 were discussed. High correlations were obtained between the cloze tests and the reading comprehension tests and curvilinear regressions between cloze tests and preference scales that confirmed the original hypotheses. These results were taken to confirm that cutting scores may be rationally determined and that algorithmically generated criterion referenced tests may be developed to measure significant educational outcomes. (Author/CTM)