Abstract
The reproducibility of bone-conduction pure-tone audiometry and speech recognition thresholds has been tested in groups of normal-hearing subjects. Each person was tested twice during the same day, and the test—retest difference was calculated. The reproducibility is presented as the standard deviation of this difference. Bone-conduction threshold measurements have a high degree of test—retest precision, whereas air-bone gaps show a large range of distribution in these normal-hearing subjects. This makes the interpretation of such gaps spurious when values are below 20–30 dB. Speech recognition threshold has the highest degree of test—retest precision of all audiometric tests, and this is probably due to the steep slope of the psychometric function at 50% intelligibility. A more detailed graphic presentation of the 50% point of intersection will bring the reproducibility down to less than 2.5 dB.