ABSTRACT
Two methods for estimating the reliability of a test (such as an achievement test) or a scale (such as an attitude scale) were covered in the previous topic: test-retest reliability, in which a test1 is administered twice, and parallel-forms reliability, in which two alternative forms of a test are administered. In both of these types of reliability, examinees must take a test twice, with a week or two intervening between the administrations. Thus, if a test is less than perfectly reliable (i.e., the scores are not perfectly consistent), which is usually the case, variables associated with time can be partly responsible. For instance, at the time of the first administration, an examinee might be emotionally upset and not do well on the test, while the same examinee might not be upset at the time of the second administration and perform much better. Such time-associated variables obviously reduce the consistency of the scores from one administration of the test to the other, reducing reliability (and the reliability coefficient).
