The sensitivity prediction from the acoustic reflex (SPAR) test was administered to 90 subjects who were classified according to four levels of mental retardation and five levels of hearing sensitivity. SPAR results showed accurate hearing level prediction for 58% of the subjects, one-category errors for 29%, and two-or three-category errors for 13%.Audiometric slope predictions were also made and were accurate for 58% of the subjects. SPAR results indicated that the greater the hearing loss, the less likely that normal hearing would be predicted. Hearing level predictions for the retarded population were of comparable accuracy but slightly poorer than those previously reported for the nonretarded. The SPAR is suggested as a useful addition to the audiologist's repertoire, particularly for retarded individuals who are difficult to test.Prediction of hearing levels from acoustic reflex thresholds (ARTs) is based on the finding that the reflex thresholds are relatively constant as stimulus bandwidth increases to a critical point, beyond which bandwidth increase produces threshold decrea~e.~. l4 For normal hearers, the difference in ARTs for broadband noise and pure tones (both measured in sensation level) is about 20 sensitivity prediction from the acoustic reflex (SPAR). Although procedural and sample differences prevent exact comparison, accurate prediction is about 60 to 70%, whereas approximate prediction (no error or one-category error) is near 95%. Jerger et a1.l' concluded from previous studies that predictions for normal hearing are very good and predictions for hearing loss, especially moderate loss, are often wrong.So far, a single study has evaluated the SPAR for retarded individuals: Niswander and Ruth13 tested 32 trainable adults using both ears. Predictions of hearing level were accurate for 57% of the ears (with a tendency to exaggerate the level), but audiometric slope predictions were accurate for only 3 1%. Although two conductive losses were missed, no sensorineural hearing loss was undetected. Niswander and Ruth13 concluded that the test was useful for predicting magnitude but not slope of hearing loss for people who are mentally retarded.In order to assess the usefulness of the SPAR with individuals exhibiting all levels of subnormal intelligence, we tested subjects showing mild, moderate, severe, and profound retardation. Moreover, we used a moderately large nuniber of subjects, and to reduce potential data correlation problems, a single ear per subject. METHODdB.3 Those with sensorineural hearing losses show smaller differences, presumably because their critical bands are wider, reducing the number of bands available for loudness s~mmation.~ Prediction of audiometric slope from ARTS is based on the assumption that a listener with a flat audiometric contour will have similar ARTS for low-pass and for high-pass filtered noise.g A sloping threshold contour supposedly attenuates the relative loudness contributions of critical bands in the higher frequencies. The result is a more sensitive ART for...