While the usefulness of branched testing over conventional paper-and-pencil testing has been in doubt, particularly for the student of average ability, this has been with reference to normative measures rather than the criterion-referenced measures characteristic of individualized instruction. A computer-assisted test model for assessing an examinee's proficiency in a set of skills for which a hierarchy of prerequisite relationships is known to exist was developed and evaluated. The test model calls for the random construction of items using item generation rules stored in the computer--an item sampling procedure that permits the test constructor to control for classification errors--and a branching strategy that tailors testing to the individual student in accordance with his competencies. Results with an individualized mathematics program showed the computer test to be highly successful in providing reliable information in substantially less time than was required by conventional methods, even though the sample included students with wide variations in competencies represented by the test unit. The reduction of time required for testing is attributed to the routing strategy rather than the item sampling procedure. (Author/MT)