Considerations for Use of Constructed-Response Items in Educational Achievement Tests

Abstract

As constructs measured by achievement tests shift from basic knowledge and skills to higher-order thinking skills, the use of constructed-response (CR) items has increased. This tendency is salient in the United States where selected-response (SR) items have been widely used, but a similar movement can also be found in Japan. The development and administration of a test would become complicated if a test consists of both SR and CR items (i.e., mixed-format). It seems necessary to enhance the theory and practice of educational measurement to handle the new challenges. In this paper, three issues concerned with the use of CR items are discussed: (1) the effect of item format on the measurement, (2) the equating of mixed-format tests, and (3) an examination of the consequences of test use. In each discussion, implications for validity and validation are emphasized. In the last section, some additional considerations are provided.