Computerized measurement of speech intelligibility. II. Normative data from a 'closed response' test.

A Danish Four Alternative Forced Choice (4AFC) test was presented to normal-hearing and cochlearly hearing-impaired listeners in two background noises: a modulated, speech-spectrum-shaped noise and an 8-persons babble noise. At the presentation to the hearing-impaired subjects the test signals were given a frequency-dependent amplification in accordance with the POGO-fitting method. Normative data are presented which indicate that 4AFC is a valid test in spite of lack of homogeneity between the alternatives allowed. On the other hand, a learning effect is demonstrated which affects the reliability. A computer-assisted analysis of the obtained consonant confusions reveals a predominance of confusions among initial consonants having the same voicing.