Predicting the understanding of Cantonese connected discourse

Abstract

Understanding of connected discourse has a high face validity in estimating an individual's ability to understand everyday speech. In situations where a direct measure of connected discourse understanding is not possible, it would be beneficial to be able to predict a person's ability to understand connected discourse from other speech and audiometric parameters. The aim of this study was to predict the understanding of Cantonese connected discourse from (1) the maximum score from the performance-intensity function of a monosyllabic word-list test, (2) the results from a lexical tone test, (3) the results from the pure tone average and the monosyllabic word-list test, (4) the results from a sentence test, lexical tone test, vowel test and consonant test, and (5) the results from pure tone average, monosyllabic word-list and lexical tone tests. Twenty-seven normal-hearing and 55 sensorineural hearing-impaired school-aged children participated in the present study. All subjects were assessed using a battery of speech tests including a connected discourse test, a sentence test, a monosyllabic word-list test, a lexical tone test, a vowel test and a consonant test. The resulting scores were subjected to a non-linear regression analysis, a standard multivariate linear regression analysis and a stepwise multiple regression analysis with the self-ratings of the connected discourse test as the dependent variable and the result(s) of the relevant testes) as the independent variable(s). Results from the present study showed that the ability of a child to understand connected discourse may be predicted from the maximum score of the performance-intensity function of the monosyllabic wordlist test, the result of the lexical tone test and the results from three different combinations of audiometric and speech measures. Furthermore, the use of anyone of the three combinations of audiometric and speech tests may give a better prediction of the understanding of connected discourse than that provided by a single speech measure. The methodology outlined in the present study may have applications in clinical situations, educational settings and cochlear implant clinics for children.