Abstract
The purpose of this three-phase investigation was to develop and evaluate a computerized adaptive test of tonal memory. In the first phase, characteristics of commercially available tonal memory tests (instrument timbre, rhythmic complexity, tonality) were systematically varied and compared to determine the types of items most likely to yield reliable and valid scores. Results from ANOVA, factor-analytic, correlation, and regression analyses indicated that synthesizer-produced, varied-rhythm tonal and atonal melodies of four to nine notes would provide reliable and concurrently valid scores over a wide range of ability levels. In the second phase, the adaptive test was constructed, and its measurement precision and efficiency were evaluated in a series of computer-simulation analyses. Results showed that the adaptive test required 5 to 11 items to yield reliabilities from .80 to .90, and 80% fewer items to surpass the reliability of the Seashore Tonal Memory Test. In the final phase, the adaptive test was field-tested on the PLATO computer system. In line with the computer-simulation findings, the adaptive test required an average of 6.05, 8.55, and 11.60 items to reach reliabilities of .80, .85, and .90, respectively.