This study compared the intelligibility of speech delivered via three transducers: a TDH-39 earphone, a Pracitronic KH 70 bone vibrator, and a Radioear 6-72 bone vibrator. CID W-1 spondees were presented to normal-hearing listeners via each transducer over a range of intensity levels.Functions relating the percentage of spondees correctly identified to stimulus level were similar for the three transducers, and notably, their slopes were comparable. This suggests that it is appropriate to use the W-1 word lists to determine speech reception thresholds (SRTs) with the KH 70 and 6-72 bone vibrators, even though these word lists were originally developed for testing by air conduction. Our subjects also identified NU-6 monosyllabic words equally well for each transducer when the words were delivered at 40 d6 above SRT. This result suggests that it is also valid to deliver the NU 6 speech test via the two types of bone vibrators.IN ADMINISTERING SPEECH TESTS via bone conduction, researchers and clinicians generally use the same procedures and speech materials as for air conduction (Barry and Gaddis, 1978;Beattie and Gager, 1980;Edgerton, Danhaurer, & Beattie, 1977). A commonly used set of materials for determining speech reception threshold (SRT) is the CID W-1 spondee words lists, originally developed by Hirsh et al (1 952) for air conduction (AC) testing. Hirsh et a1 determined that for normal hearing listeners, the function relating the percentage of spondees correctly identified to the intensity level of the speech (i.e., the performanceintensity or articulation function) rose vcry steeply for scores between 20 and 80%. This implied that an appropriate operational definition of threshold is the intensity level at M. hich the listener correctly identifies 50% of the words. This is because in this range of percentages, small changes in the level of the speech result in relatively large changes in the score. Consequently, SRT can be determined with a high degree of accuracy and reliability (Beattie, Forrester, & Ruby. 1977: Chaiklin andVentry, 1982).However, the data of Hirsh et a1 ( 1952) were obtained with earphones or loudspeakers. These transducers have broader frequency responses than do bone vibrators. For examplc, the output of the TDH-39 earphone extends to about 7 to 8 kHz before beginning to drop off, whereas the Radioear B-72 bone vibrator has little output above 4.5 kHz (see Figs. 1 and 2). This would be expected to result in differences in the frequenc) content of speech signals arriving at the cochlea. Further differences in the spectrum of speech at the cochlea would be expected because the routes of transmission of the air and bone conducted signals are not the same (Bekesy, 1960). Given these discrepancies, it is possible that performance-intensity functions for spondees delivered via bone may differ from those obtained with air conducted signals. For example, the slope of the function could be lower in the case of bone conducted signals. If this were so, the criterion scorc for S R r used in AC test...