Speech discrimination in the presence of competing cafeteria and speech babble noise was measured in elderly subjects with bilaterally symmetrical sensorineural hearing losses. All demonstrated at least one central or retrocochlear audiometric sign. The purpose of the study was to investigate whether low-frequency attenuation, dichotic or diotic presentation of speech in noise would improve speech discrimination in this population. Previous work by Pickett and associates using synthetic formants, had demonstrated release from low-frequency masking in some younger subjects as a function of these conditions. An as yet unpublished study by Kaplan and Pickett demonstrated similar results in presbycusics. It was hypothesized that comparable release from low-frequency masking would occur if speech in noise were presented under attenuated low-frequency, dichotic and diotic conditions. Speech discrimination was measured under earphones with noise and speech in the preferred ears under the following conditions: unaltered monotic, and monotic with attenuation of the band below 1000 Hz by 5, 10, and 15 dB; in addition, dichotic (band split) and diotic. For the dichotic condition, the band was split at 1000 Hz with filter skirts of 55 dB per octave. The high-pass band was delivered to the preferred ears. Modified Rhyme Test stimuli in open and closed set formats and SPIN Test sentences were used. For the group as a whole, dichotic and diotic conditions were significantly superior (0.05 and 0.01 levels) to unaltered monotic. When subjects were grouped according to degree of loss, the mild, loss group showed greater dichotic-monotic and diotic-monotic superiority than the moderate loss group. When subjects were grouped into “SSI Normal” and “SSI Abnormal” categories (two different types of central involvement), based on performance on the Jerger Synthetic Sentence Inventory (SSI) in ipsilateral competing message, there were no significant differences between groups, except for larger dichotic-monotic differences for the SSI Abnormal group. [Work supported by NINCDS.]