Recording and evaluation of an American dialect version of the Four Alternative Auditory Feature test

The Four Alternative Auditory Feature test (FAAF) is a word-based closed-set speech recognition test. Because the original test materials were recorded in British English dialect, it is not appropriate for use in the United States.

The purpose of this study was to produce an American dialect FAAF (AFAAF).

The AFAAF materials spoken by a native American-English speaking male were recorded and digitally edited. In the validation study, the AFAAF was administered monaurally at five signal-to-noise ratios (SNRs) in both ears for each listener.

A total of 20 young adults with normal hearing participated in the validation study.

For each participant, speech recognition scores were collected in one session. The speech level was fixed at 70 dB SPL and the steady-state talker-matched noise level was varied, resulting in five SNRs from -15 to -5 dB. One full list (80 words) was used for each SNR. For each participant, a performance-intensity (PI) function was fit to the discrete mean percent correct scores for the five SNRs according to a best-fit, three-parameter sigmoid function. In addition, scores for the left and right ears were compared to examine test-retest reliability.

RESULTS show that the slope of the PI function is 6% per dB, the mean test-retest difference scores for the five SNRs are within 3 rationalized arcsine units (rau), and the 95% critical difference for the 80-word scores is 12 rau. Compared with the FAAF, the slope of the PI function for the AFAAF is slightly less steep. Test-retest reliability of the AFAAF is at least equal to that of the FAAF.

It is concluded that the AFAAF is similar but not identical to the FAAF. The AFAAF is now available for measuring speech recognition performance in listeners who use American English as a native language.