Statistical models for predicting Edgerton-Danhauer NST scores from pure-tone thresholds

Pure-tone thresholds (PTTs) and phonemic-scored responses to the Edgerton and Danhauer (1979) Nonsense Syllable Test (NST) at 25, 35, 45, and 55 db SL re each S's SRT, were taken from 97 Ss of 4 earlier studies of Danhauer and colleagues. All studies had used the same stimuli and procedures. Among the 97 Ss were heterogeneous subpopulations of normally-hearing and sensorineural hearing-impaired Ss, aged 6-81 yrs. PTTs at octaves from .25-8 kc/s, singly and in various weighted combinations, were explored by step-wise multiple linear regression analyses for the best equation to predict an individual's NST score at each SL. Included were the two weighted PTT averages (over .5, 1, and 2 kc/s) proposed by Butts, Schoney and Ruth (Asha, 1984, 26, 64). PTT2 kc/s was the single most predictive datum. The most predictive equation at one NST SL was not necessarily the best at another SL. At each NST SL, the most predictive equation was always one or other of the equations of Butts et al, and at the higher levels the best equations incorporated in addition certain weighted PTTs at .5 and 1 kc/s. The results indicated that it is possible to predict a S's NST scores from PTTs with reasonable accuracy: from 86-91% of Ss' NST scores were predicted within +/- 10%. Further validation and refinement of these models on larger and different subject samples may result in equations predicting NST scores from PTTs with those difficult-to-test patients in whom speech recognition capacity cannot well be tested directly.