“On the Adequacy of Standardized Tests Administered to Extreme Norm Groups”, 1951 ():
Most standardized tests are recommended by their publishers for use in more than one grade. Frequently, some convenient grouping corresponding to a prevalent type of school, such as the senior high, is suggested in the manual of directions. Quite a few tests are recommended for an even wider range, this being particularly true of intelligence scales. Thus presumably the Otis Quick-Scoring Mental Ability Test, Gamma Test, is equally useful anywhere from Grade 9 through Grade 16, while the California Test of Mental Maturity, Advanced Form, is designated for Grade 9–adult.
Thurstone found that “the factorial content of a test will change as it is given to populations that differ in age and schooling” (14, p. 43), and common sense long ago told us that IQ’s based upon a children’s test administered with a shortened time limit to adults probably do not have the same meaning as they would for fifth graders…
[test results on various groups]
…In order to make their tests more salable, a considerable number of authors have recommended them for use in grades below or above those for which the tests were initially designed. Thus questions concerning changing factorial content and difficulty level arise. As an illustration of a test too hard for the lowest grade suggested by its constructors, the writer arbitrarily selected the Nelson-Denny Reading Test for Colleges and Senior High Schools, on which there was data available. This instrument was found to be of unsuitable difficulty for the lower half of a typical ninth grade (161 pupils) in a New England public coeducational senior high school. During the analysis several negative reliability coefficients were secured. This statistical anomaly and theoretical issues related to it are discussed briefly.