the talker (Lloyd and Price, 1971), visual acuity (Hardick, Oyer, and Irion, 1970), knowledge of the language, live versus recorded materials, and degree of hearing loss (Erber, 1979). For example, word recognition performance improves 19-28 percent for individuals with severe hearing loss with the addition of visual cues, but improves only 1-15 percent for individuals with profound hearing loss (Erber, 1975). Although reception of everyday sentences is usually optimal for audition + vision conditions (Tye-Murray, 1998), some individuals with severe or profound hearing losses may exhibit very poor performance (10 percent correct) for recognition of everyday sentences presented with auditory + visual cues (Sims and Hirsh, 1982). These unaided performance levels do not reflect the combined benefit of speech-reading and use of amplification or a cochlear implant in suitable candidates.

Multicultural and Multilingual Issues in Evaluation of Speech Recognition

U.S. society is becoming increasingly multicultural and multilingual. Individuals seeking audiometric evaluation may have no knowledge of the English language or may have limited fluency in English. Presentation of a standardized English speech recognition test to these individuals is problematic for several reasons. First, a lack of familiarity with the vocabulary is known to reduce performance on a speech recognition test. As a result, nonnative speakers of English obtain lower scores on English speech recognition tests than do native speakers of the language (Gat and Keith, 1978). Second, listeners whose first language is not English perceive individual consonant and vowel phonemes differently than native speakers of English (Danhauer, Crawford, and Edgerton, 1984). Finally, nonnative speakers derive less meaning from sentence-length materials than native speakers of English, in part because of differences between the overall rhythmic pattern of English and that of many other languages. For bilingual speakers, the age of second language acquisition is an important factor influencing proficiency in English. This is particularly apparent on speech recognition tests in noise: nonnative adult speakers of English who learned English before age 6 perform better in noise than adult listeners who learned the language after puberty, even though all listeners achieve nearly perfect performance in quiet (Mayo, Florentine, and Buus, 1997).

Despite these obstacles, it remains desirable to evaluate a listener’s speech recognition performance during an audiometric assessment. A number of alternative materials and methods have been recommended for evaluating nonnative speakers of English. The preferred strategy is to present a speech recognition test for which recordings are available in

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement