SPEECH INTELLIGIBILITY OF PEDIATRIC HA USERS 323 RESEARCH ON SPOKEN LANGUAGE PROCESSING Progress Report No. 23 (1999) Indiana University Speech Intelligibility of Pediatric Hearing Aid Users1 Mario A. Svirsky, Steven B. Chin, Matthew D. Caldwell and Richard T. Miyamoto2 DeVault Otologic Research Laboratory Department of Otolaryngology-Head & Neck Surgery Indiana University School of Medicine Indianapolis, IN 46202 1 This study was supported by NIH-NIDCD grants DC00064 and DC00423. Terri Kerr assisted with data management; Allyson Riley, Amy Robbins, and Susan Sehgal collected the data, and Karen I. Kirk and Susan Sehgal gave useful suggestions for the manuscript. All this help is gratefully acknowledged. 2 All of the authors are also affiliated with the Speech Research Laboratory, Department of Psychology, Indiana University, Bloomington, IN 47405.SVIRSKY, CHIN, CALDWELL AND MIYAMOTO 324Speech Intelligibility of Pediatric Hearing Aid Users Abstract. This study examined the speech intelligibility of profoundly, prelingually or congenitally deaf children who use hearing aids. Children were one to fifteen years old and they were classified into subgroups according to residual hearing (PTA between 90 and 100, 100 and 110 or greater than 110 dB HL) and communication mode (either oral or total communication). They read lists of standard sentences which were played back to panels of three “naive” listeners who were not familiar with the speech of the deaf and who did not know which subgroup the children belonged to. The data revealed a strong significant trend toward higher intelligibility for children with more residual hearing, and a significant trend toward higher intelligibility for users of oral communication than for total communication users. However, the latter trend was much more pronounced for some ranges of residual hearing than for others, and it may have been partly due to a sampling effect. A third trend showed significantly higher intelligibility levels at older ages, but this was particularly pronounced for children with PTA’s between 90-100 dB HL, and for the majority of oral communication users (and only a few total communication users) with PTA’s between 100-110 dB. These results suggest that the amount of residual hearing (possibly in interaction with the communication mode used by the child) may be an important factor in the development of intelligible speech. Introduction A connection between prelingual hearing impairment and atypical speech production has been observed for millennia (e.g., Hippocrates, 1853, p. 609), and it is known that the speech of prelingually deafened children can be very difficult to understand. Nevertheless, throughout history, people have attempted, with varying degrees of success, to impart to children with hearing impairment the skills necessary to allow them to communicate successfully in a world so dependent on spoken language. Among the myriad measures used to assess speech production in children (e.g., phonemes correct, suprasegmentals correct, aerodynamic deviance), the one with the highest face validity as regards the need to communicate using spoken language is overall speech intelligibility. Measures of overall speech intelligibility cut to the quick, addressing the important question, “Can this child be understood?” The question of whether a child with hearing impairment can be understood appears a simple question, but stating the question, and, perhaps more importantly, answering the question, do not by any means lack scientific subtlety. The question “Can this child be understood?” immediately raises obvious follow-up questions: “By whom?” and “How much of the time?” and “At what level?” The scientific literature on speech intelligibility among children with hearing impairment is not uniform in its approaches to these questions, and how best to address the question of speech intelligibility can be as much a matter of practical considerations as of theory. Measuring the speech intelligibility of children with hearing impairment has generally relied on two types of tasks: (1) rating scales and (2) write-down procedures (Samar & Metz, 1988). In the first type of task, listeners make explicit judgments about the talker’s (overall) speech intelligibility by assigning numerical values to samples of speech (e.g., the NTID rating scale: Subtelny, 1977), whereas in the second type of task, listeners “write down what they thought each child said” (Monsen, 1981). As Metz, Schiavetti, & Sitler (1980) point out, write-down procedures appear to have higher face validity than a rating task, because results depend on the listener actually and literally understanding what has been said. Moreover, write-down procedures are relatively insensitive to vocal qualities of speech; theseSPEECH INTELLIGIBILITY OF PEDIATRIC HA USERS 325 may contaminate responses on a rating scale, but they need not of necessity degrade message intelligibility (Samar & Metz, 1988). On the other hand, write-down protocols are time consuming and labor intensive and therefore expensive (but see Samar & Metz, 1988), whereas rating tasks are viewed as relatively “quick and easy” (Metz et al., 1980). Some considerations making write-down intelligibility assessments difficult to administer are that listeners must respond to every word heard, and that individual listeners should not hear the same word or sentence more than once, to avoid an order effect from increased familiarity with the spoken material. Disadvantages of rating scale tasks are the relatively lower face validity of the protocol, as noted above, and the lack of material for phonemic analysis of listener misidentifications. A second consideration in speech intelligibility assessment is the type of listener judge, roughly, experienced vs. inexperienced (naive). “Experienced” generally means experienced with the speech of people with hearing-impairments, and such experienced listeners may include educators of the deaf, audiologists, and speech-language pathologists (e.g., McGarr, 1983). Naive listeners are those who have had little or no experience with the speech of the hearing impaired. There is often a connection between the type of task involved in speech intelligibility assessment and the type of listeners involved in the protocol. Rating scales, especially equal-interval scales, are often administered to experienced listeners (e.g.,