Speech audiometry in noise-exposed workers: the SRT-PTA relationship revisited

The speech recognition threshold (SRT) is believed to be related primarily to the pure-tone average (PTA) and the steepness of the hearing loss. However, there are indications that it may also be influenced by perceptual or cognitive-linguistic factors, or both, such as meaningfulness of the speech stimuli. The purpose of the present study was to ascertain the correspondence between SRT and PTA in noise-exposed workers with various degrees of speech recognition threshold shift in noise. To this end, a total of 807 SRTs and PTAs collected from fluent speakers of Quebec French noise-exposed workers were compared. Measurements of context effects on speech recognition were taken based on a general hypothesis that they should be facilitating phoneme or word restoration in conditions of high stimulus uncertainty as present in SRT assessment, thus acting to confound the SRT PTA relationship. Using principal components analysis, we found a significant effect not only of low-frequency hearing sensitivity but language context effects on SRT. After a correction was introduced to partial out these linguistic context effects, correlations between SRT and PTA increased but they were lower than predicted. In a related treatment analysis, we found a large number of observations (230 out of 807) where SRTs were more sensitive than PTAs by a factor of 8 to 16 dB. This was the case even though correlations between the two measurements were within the range commonly advocated in the field of clinical audiology (0.85-0.95). This was interpreted as a sign of phonological and lexical context effects on the speech recognition task actually used by individual subjects to facilitate speech understanding, to the point perhaps of making it as simple as the detection of pure tones.