How to Achieve
Normal Speech-in-Noise Perception
for Learning-Disabled Children
Popular version of paper 4aSCa15
Presented Thursday Morning, December 7, 2000
ASA/NOISE-CON 2000 Meeting, Newport Beach, CA
Some learning disabled children have speech perception difficulties. In particular, these children have great difficulty discriminating speech sounds that are minimally distinct. This impairment is believed to become particularly troublesome when speech perception must take place under noisy listening conditions, such as are likely to be encountered in a typical classroom.
In this study, we wanted to see whether the abilities of these children to perceive speech in noisy listening environments could be overcome by simply having the talker speak clearly. The rationale behind this test came from previous studies that have shown that hearing impaired adults derive great speech perception benefit from naturally produced "clear speech." Specifically, on tests that measure speech perception accuracy in terms of the percentage of words in an utterance that are correctly perceived, hearing impaired adults typically show a 17-20% improvement when the talker switches from a conversational to clear speaking style.
Furthermore, most (if not all) talkers are able to produce some degree of speech clarity by simply being instructed to "speak as if the listener is hearing impaired." That is, talkers can spontaneously go into a clear speech mode, with no prior training, resulting in dramatic speech perception improvements for hearing impaired adults. Given recent interest in the use of digital enhancement techniques in speech and language training procedures for learning disabled children, we wanted to investigate whether naturally produced clear speech yields perception benefits for this population. We reasoned that if learning impaired children, like hearing impaired adults, can be shown to derive significant benefit from this simple and natural talker-based modification then we may have found a very straightforward means for enhancing the speech perception, and ultimately the learning abilities, of these children.
We tested the speech perception accuracy of a group of 115 school-aged children when presented with sentences that were digitally mixed with broad-band noise. Of these children, 77 had been clinically diagnosed prior to entry into the study as having either a learning disability, attention deficit disorder or both. The remaining 38 children had no history of learning or attention problems and scored within normal limits on all tests in a study-internal psychoeducational test battery. All children had normal hearing and normal intelligence.
The test sentences were selected from a set of standardized sentences that were designed for pediatric audiometric testing and included only words that are highly familiar to young children. Recordings of 64 such sentences were made from two talkers who produced each sentence in two conditions. In the "conversational" condition the talkers were instructed to speak at their normal pace, as if the listener were someone highly familiar with their voice and speech patterns. In the "clear speech" condition, they were told to speak with extra care, as if addressing a listener with a hearing loss or from a different language background.
Listen to sample of Female Clear Condition
Listen to sample of Female Conversation Condition
Listen to sample of Male Clear Condition
Listen to sample of Male Conversation Condition
At the time of testing, the recordings were mixed with broad-band noise and presented to the children over a loudspeaker. The child's task was to listen to the sentence and then to repeat whatever she or he heard to the experimenter, who recorded the number of keywords that the child repeated correctly. For each child, half of the sentences were produced in the conversational speaking style and half were produced in the clear speaking style. Furthermore, within each of these two style conditions, half of the sentences were presented with a higher level of noise and half were presented with lower level of noise (-8 dB and -4 dB signal-to-noise ratio, respectively.) Each child listened to the sentences as produced by just one talker.
Three major findings emerged from this study (summarized in Figure 1). First, on average (across both talker, noise level and speaking style conditions) the group of impaired children had poorer sentence-in-noise perception accuracy than the group of normal children. This basic finding provides empirical support for the widely held belief that the impaired speech perception abilities of these learning disabled children extends to impaired perception of sentence length utterances when presented in noise.
Second, both groups of subjects showed a dramatic improvement in speech perception accuracy for the clear speech sentences over the conversational speech sentences. The average improvement for both the impaired and normal children was about 10%. The third finding was that the average sentence-in-noise perception score for the group of impaired children when presented with sentences produced in clear speech reached the same level of performance as for the group of normal children when presented with sentences produced in conversational speech. In other words, the talker-based speaking style manipulation was sufficient to bring the speech perception abilities of the impaired children within normal limits.
These results have significant implications for the enhancement of speech communication with these children. Until we have a full understanding of the underlying causes of the speech perception deficit exhibited by these children, it is difficult to devise efficient, effective and scientifically-grounded training techniques that can provide a "cure" for the problem.
Nevertheless, as shown by this study, there is a very straightforward modification on the part of the talker that can quite effectively bring the speech perception performance of impaired children within normal limits. In this study, the vast majority (83%) of the learning impaired subjects derived some benefit from this talker-related modification, and over half (57% of the 77 impaired children) improved their performance by over 7.5%, which is the average difference between the impaired and normal children's performance level on this task. In other words, the improvement in speech perception accuracy achieved by this naturally produced talker modification is consistent across subjects, immediate and essentially cost-free.
The key is for the talker to be aware of the listener's problem and, despite the fact that the listener appears to have normal hearing acuity, the talker should adopt a speaking style that one might naturally adopt for a (more obviously) hearing impaired listener. That is, when talking to a child with a learning impairment, one should take care to speak clearly. It really does help.
[This work was supported by NIH-NIDCD grants DC 03762 and DC 01510.]