Diagnostic precision of open- versus closed-set word recognition

Abstract

The precision of forced-choice (closed-set) and open-ended (open-set) word recognition (WR) tasks for identifying a change in hearing was examined. WR performance for closed-set (4 and 6 choices) and open-set tasks was obtained from 70 listeners with normal hearing. Speech recognition was degraded by presenting words from the Modified Rhyme Test in speech-shaped noise (-8, -4, 0, and 4 signal-to-noise ratios) or by processing words using a sinewave vocoder (2, 4, 6, and 8 channels). The results for the two degraded listening conditions yielded similarly shaped, monotonically increasing psychometric functions. The closed-set tasks had shallower slopes and higher scores than the open-set task for the same condition. Fitted, average psychometric functions were the input to a computer simulation conducted to assess the ability of each task to identify a change in hearing. Individual data were also analyzed using 95% confidence intervals for significant changes in scores for words and phonemes. These analyses found the following for the most to least efficient condition: open-set (phoneme), open-set (word), 6-choice closed-set, and 4-choice closed-set. Greater than an order of magnitude more trials were needed for the 4-choice condition to equal the precision of the open-set word condition scored by the percentage of correct phonemes.The precision of forced-choice (closed-set) and open-ended (open-set) word recognition (WR) tasks for identifying a change in hearing was examined. WR performance for closed-set (4 and 6 choices) and open-set tasks was obtained from 70 listeners with normal hearing. Speech recognition was degraded by presenting words from the Modified Rhyme Test in speech-shaped noise (-8, -4, 0, and 4 signal-to-noise ratios) or by processing words using a sinewave vocoder (2, 4, 6, and 8 channels). The results for the two degraded listening conditions yielded similarly shaped, monotonically increasing psychometric functions. The closed-set tasks had shallower slopes and higher scores than the open-set task for the same condition. Fitted, average psychometric functions were the input to a computer simulation conducted to assess the ability of each task to identify a change in hearing. Individual data were also analyzed using 95% confidence intervals for significant changes in scores for words and phonemes. These analyses ...