| Principles of item response theory (IRT) underlie this computer-adaptive listening-comprehension test developed to assist in placing students in beginning-level French courses. This dissertation documents the construction and validation of the test which took place in four phases: (a) defining the listening-comprehension construct; (b) conducting a domain study and preparing a test blueprint and test specifications; (c) administering an initial test, calibrating the test items, conducting an item analysis, selecting items for the test bank, and determining cut scores; and (d) collecting evidence of and building a case for validity.;The listening-comprehension construct is defined in terms of three theoretical cognitive behaviors: attending to aural stimuli, identifying or decoding lexical items and their underlying propositions, and interpreting or assigning meaning in the context of the listener's purpose. The definition also acknowledges the important role of inferencing in all three cognitive behaviors.;The domain study identified text, listener, and task characteristics relative to the first three semesters of Brigham Young University's French program. In particular, the study showed that 83 percent of classroom listening is interactional (students seeking to recognize a personal component in utterances) and that students rarely play the role of overhearer in the classroom.;The initial test, consisting of 56 items, was administered to 265 students. After an item analysis based on classical-test theory and IRT methods, 54 test-bank items were calibrated for use in the computer-adaptive test. Cut scores were determined using a contrasting-groups method.;The argument for validity, organized according to Samuel Messick's six aspects of construct validity, showed (a) The test is relevant to and representative of the target domain; (b) There is evidence of unidimensionality and minimal guessing; (c) The assumption of local independence was met; (d) The cut scores are useful and valid, and test results can be generalized across administrations to the target population; (e) The test results correlate with logically related constructs and external variables, such as test scores on grammar and reading comprehension, years of high-school French, and self-confidence in listening-comprehension ability; and (f) The consequential aspect of validity supports test use. |
