Further investigation of vowel items in multiple-choice speech discrimination testing.

Abstract

A list of 25 vowel-test items employing a multiple-choice (or closed-set) response system was presented to 94 patients with hearing impairments and 10 with normal hearing listening through a low-pass filter. The items had been demonstrated empirically to be the most difficult of 288 similarly constructed items. Even these 25 items produced relatively few errors in the two groups. No differences occurred in probability of error on individual phonemes among three types of hearing impairment (Meniere’s disease, presbycusis, and noise-induced loss). Some slight evidence suggested, however, that the kinds of error responses made for /ε/ presentation were somewhat more varied among patients with Meniere’s disease than among those with presbycusis or noise-induced loss. In general, multiple-choice items structured to permit confusion of one vowel with another failed to show promise as good test items for speech discrimination.