Psychometric Equivalence of Recorded Spondaic Words as Test Items

Abstract

In the determination of the speech-reception threshold (SRT), spondaic words are assumed to be homogeneous with respect to intelligibility; and the assumption of equal intelligibility requires that the words be comparable for all signal levels. Previous attempts to assess the equal intelligibility assumption using word thresholds as the sole criterion are not an adequate basis for specifying the equality of intelligibility. In the present study, the recorded spondaic words (Tillman recording) were analyzed in an attempt to create a more homogeneous set of spondaic words for future laboratory work. To achieve this goal, the data reported by Young, Dudley, and Gunter (1982) and data collected in our laboratory were fitted to a logistic function (psychometric function) from which a 50% point (threshold) and slope were obtained. To specify their acoustical parameters, the recorded spondaic words were digitized and the RMS level and duration of each syllable and word were calculated. None of the RMS or duration measures were correlated with word thresholds, so no attempt was made to equate level or duration. On the other hand, when the threshold of each word was adjusted to equal the mean threshold of the set ( n = 36), the dispersion among word thresholds and slopes was greatly reduced. Further, we recommend that small sets of "equally intelligible" spondaic words not be used for clinical testing because set size is a strong factor in determining threshold for spondees (Meyer & Bilger, 1997; Punch & Howard, 1985).