Standard Error Measurement in Speech Discrimination Testing

Abstract

Speech discrimination data obtained by sequential testing with 12 normal-hearing listeners were described by Bode and Carhart at the 80th Meeting of the Society in Houston [J. Acoust. Soc. Amer. 49, 95(A) (1971)]. Group results showed that sequential testing provided discrimination functions essentially identical to those measured by conventional constant stimulus methods. Test-retest effects were small and quantifiable. We now have analyzed our results in terms of individual performance and will report these data for scores obtained both within and between test sessions. Of particular interest to us is the fact that a standard error (SE) can be calculated for individuals with sequential testing, even using only one test. Such a determination is not possible with orthodox discrimination test procedures. There are many circumstances, particularly in clinical audiology, in which an investigator or clinician would like to have both a rapid efficient estimate of the SE associated with his measure and one which can be calculated for an individual during a single test administration. Such a calculation would allow the examiner to reach conclusions, for any given obtained score, regarding the probable limits of corresponding true scores. These estimates for individuals, together with those based on group performance, will be described and contrasted. [Research supported by National Institute of Neurological Diseases and Stroke.]