8th Speech in Noise Workshop, 7-8 January 2016, Groningen

Long-term effects of hearing impairment for word processing and speech recognition in noise

Rebecca Carroll(a), Anna Warzybok(b), Esther Ruigendijk
University of Oldenburg, DE

(a) Presenting
(b) Attending

Speech recognition in noise has been proposed to correlate with a listeners’ vocabulary size (McAuliffe et al., 2013; Benard et al., 2014). One reason for such a relation is that a larger lexicon requires a more detailed representation in the mental lexicon to successfully distinguish phonological neighbors. Some researchers (Arlinger, 2003; Classon et al., 2013) have suggested untreated long-term hearing loss to lead to cue deprivation that may render detailed lexical information (e.g., phonetic details) less reliable. Carroll et al. (submitted) showed age effects of lexical access times and speech recognition scores in younger versus older listeners with normal hearing. Measures of vocabulary size and working memory did not show group effects. Effective lexical access time was associated with speech recognition scores. Lexical access time depends on the size of the mental lexicon and the relative efficiency with which a lexical candidate is found when boundaries between phonological neighbors become less distinct due to phonetic blurring caused by impaired hearing. We therefore asked whether hearing impaired listeners show long-term effects regarding the mental lexicon that relate to speech perception in noise. We tested 22 older listeners with normal hearing (60-78 yrs.) and 20 older listeners with mild-to-moderate hearing loss (60-80 yrs.). We measured speech recognition thresholds (SRT) in noise correct using a German everyday sentence test with adaptive procedure to establish the individual SNRs yielding 50% and 80% correct recall. Speech was presented with NAL-R and without amplification. Both groups also completed a small battery of selected individual differences measures with a focus on the mental lexicon. Data collection of hearing impaired listeners is ongoing. First data show strong differences of about 4 dB SNR in speech recognition scores without amplification. With NAL-R amplification, SRTs yielding 50% correct were comparable, with only 1 dB difference, but SRTs yielding 80% correct still showed a 3 dB benefit for the group with normal hearing. Whereas measures of vocabulary size were comparable between groups, working memory showed a small disadvantage for hearing impaired listeners. Hearing impairment was also associated with much longer lexical access times. The combination of slow lexical access, comparable vocabulary size and somewhat lower verbal working memory is expected to contribute to relatively poorer speech recognition scores.

Last modified 2016-05-12 14:22:09