Evaluation of two clinical versions of the articulation index.

The present study evaluated the predictive accuracy of two clinical versions of the articulation index (AI). AI calculations were made and word recognition scores obtained for normal-hearing young adults and elderly hearing-impaired listeners in quiet and in a noise background. The impaired listeners were also evaluated for both aided and unaided listening conditions, with the aided AI values adjusted by the measured real ear insertion gain for each listener. Results revealed that the two clinical AI procedures yielded virtually identical values, and both were accurate predictors of speech recognition performance for unaided conditions. Aided speech recognition performance was predicted less accurately by both methods, especially when a noise background was present. On the other hand, within a given listener, both AI methods were valid indicators of the relative performance across listening conditions.