A technology for domain or criterion referenced testing was developed to evaluate the Minnesota Mathematics and Science Teaching Project (MINNEMAST) in grades K-6. The venture was influenced by the low achievement of MINNEMAST students on standardized tests. After instructional objectives were selected and operationally defined, corresponding test items for each course unit were developed. As the collection of items grew, they were arranged in formal tables called item forms. Item forms served two basic purposes: they obviated the need to store individual items, by substituting a set of written rules through which items could be generated when needed; and they enabled the relationships among items to be traced by specifying item characteristics, including stimulus and expected response, and instructions for administering and scoring. The test was actually an ordered set of items administered individually according to a matrix sampling design. Because each child received different item sets, there were as many tests as children sampled. Results were interpreted in terms of patterns of responses to item sets with related stimulus and response characteristics. (Recommendations for curriculum development and evaluation, sample item forms, and a sample interpretation of test results are included.) (CP)