Hearing disability can be measured by questionnaires, interviews or self-ratings. As these measurement techniques are not always appropriate for direct application in individual cases, particularly in the context of compensation claims, the assessment is sometimes reduced to a surrogate measure derived from audiometric thresholds. No final agreement emerges from correlation studies on hearing disability as to the optimal set of audiometric descriptors. However, a multiple regression equation, describing the relationship between audiometric data and numerically expressed self-ratings of disability, provides a means for predicting disability with relatively good precision, using some index derivable from the audiogram. The aim of the present study was to consolidate or amend current prediction schemes based on tonal audiometric data by investigating the problem of differential bilateral weighting of threshold frequency through multiple regression performed on experimental data coming from an unbiased group of hearing impaired subjects. In addition to tonal thresholds, this corpus included simple, quantitative self-ratings for disability. The sample was considered to be a representative cross-section of adult hearing impaired ENT patients. The results indicate that prevailing bilateral weightings are likely to overestimate the importance of the worse ear. The regression results did not incorporate the usual threshold values and frequency weightings. The outcome nevertheless offered a multiple correlation coefficient that was higher than most correlations reported in the literature between disability ratings and threshold values.