Content and procedural learning in repeated sentence tests of speech perception

Repeated testing of speech perception is unavoidable in evaluating the benefits of hearing aids and auditory rehabilitation, but procedural and content learning due to repeated test administration can masquerade as a general improvement in speech perception. A previous study of the speech reception threshold (SRT) in quiet reported procedural learning that was sufficiently large to call into question the use of repeated sentence testing in evaluating the effects of auditory rehabilitation. The objective of the first experiment was to measure the effects of content and procedural learning in noise using the Hearing in Noise Test (HINT) when some sentences were repeated and others were not. The objective of the second experiment was to estimate the effects of procedural learning in a larger group of listeners using both the HINT and the Quick Speech in Noise test (QuickSIN) without sentence repetition across test sessions. The objective of the third experiment was to evaluate content learning in the HINT and the QuickSIN when sentence tests were repeated at intervals of several months.

In experiment 1, eight normal-hearing listeners completed five 1-hr test sessions on separate days. All sessions included sets of HINT sentences that were presented twice per session to evaluate content learning. Sessions 1 and 5 also included sets of unique sentences to measure procedural learning. In experiment 2, 23 young normal-hearing listeners completed three sessions over a 10-day period with unique HINT and QuickSIN sentence lists presented in each session. In experiment 3, 11 older, normal-hearing listeners completed three sessions of unique HINT and QuickSIN sentence lists, as in experiment 2. After an interval corresponding to a course of auditory rehabilitation training or hearing-aid acclimatization, the listeners were tested with the same sentence lists.

In experiment 1, the SRT for repeated sentences improved by an average of 2.7 dB, whereas that for unique sentences showed an insignificant 0.3 dB change. These results demonstrate that HINT SRTs can be affected by content learning for repeated sentences, but are minimally affected by procedural learning for unique sentence material. Significant procedural learning was found only in the first session. In experiment 2, HINT SRTs improved by 0.2 dB per session whereas improvements on the QuickSIN (0.1 dB per session) failed to reach statistical significance. In experiment 3, both tests showed significant improvements; HINT SRTs improved by 0.5 dB and QuickSIN SRTs by 0.4 dB.

Both the HINT and the QuickSIN provide stable and sensitive measures of speech perception across repeated test sessions provided that sentences are not repeated. Practice with at least two sentence lists is needed to eliminate the initial effect of procedural learning in the first session. The results with the HINT and QuickSIN at moderate noise levels differ from previous results of sentence testing in quiet, whereas SRTs improved 6 to 9 dB over five sessions for both repeated and unique sentences. Differences between testing at moderate noise levels and in quiet seem to account for the difference in the stability of these sentence-test measurements.