Aspects of generating syntax-free word lists for quantifying speech intelligibility: A pilot study

Several possibilities exist for quantifying speech intelligibility in noise, for example tests with single words or sentences. In the clinical routine, usually a single sentence test is used repeatedly, so that its vocabulary is often known to the patients. Thus, not only speech intelligibility but also long-term memory is tested. New speech material is then required to explicitly address intelligibility. With the aim of providing new and optimized material, in this pilot study, semantically, syntactically, and phonetically balanced words were taken from German sentence tests, based on databases for written and spoken German words. The resulting vocabulary consists of 54 words (nouns, adjectives, verbs, and numbers), for which word-reception thresholds were determined for ten normal-hearing subjects. With these thresholds, five-word lists of equal semantic context, equal word intelligibility, and with balanced phonetic cues were built. Non-acoustic advantages of an intelligibility test constructed this way are phonetic balance and semantic coherence without syntactic context, which aim at preventing syntactic and phonetic cues from affecting speech-intelligibility thresholds. As an initial verification, list intelligibilities were measured with the above subject sample. The results of this preliminary pilot study show no systematic difference for consonants and vowels. This indicates that the lists were successfully constructed without phonetic biases.