The author studied the frequency filtered speech audiometry to evaluate the index of differential diagnosis of central deafness and to know which frequency filtered speech word is proper for differential diagnosis. The maximal discrimination score, speech reception threshold(SRT) and pure tone threshold were analyzed in 86 normal hearing 64 conductive deafness, 60 sensorineural deafness, 38 cochlear deafness and 22 retrocochlear deafness using Korean normal word and frequency distorted speech word such as 1,700 Hz high pass filtered word and 1,200 Hz low pass filtered word lists. And then following results were obtained : 1) In normal hearing cases, maximal discrimination score(MDS) with 1,700 Hz high pass filtered speech audiometry was 96.09% and that of 1,200 Hz low pass filtered speech audiometry was 81.23%. It means that 1,200 Hz low pass filtered speech word is more sensitive than 1,700 Hz high pass filtered speech word. 2) The MDS and SRT with 1,200 Hz low pass filtered speech audiometry were worse than 1,700 Hz high pass filtered speech audiometry in each deafness. The sensitivity of 1,200 Hz low pass filtered speech audiometry was higher than 1,700 Hz high pass filtered speech audiometry. 3) In sensorineural deafness, cochlear and retrocochlear deafness, the gap between MDS of normal speech audiometry and that of frequency filtered speech audiometry was 2 times higher in 1,200 Hz pass filtered speech audiometry than in 1,700 Hz high pass filtered speech audiometry. 4) The gap between speech reception threshold of ordinary speech audiometry and that of frequency filtered speech audiometry was 2-5 times higher in 1,200 Hz low pass filtered speech audiometry than in 1,700 Hz high pass filtered speech audiometry. Author concludes that 1,200 Hz low pass filtered speech audiometry is superior to 1,700 Hz high pass filtered speech audiometry for differential diagnosis of retrocochlear deafness.