To investigate the feasibility of the round window stimulation electrical evoked auditory brainstem response (EABR) test, and optimize the parameters of recording and stimulation electrodes positions.Ten healthy Hartley guinea pigs (20 ears) were used for the EABR test. The positive stimulation electrodes were placed into the round window niche, the animals were divided into three group according to the negative electrodes position, group A: the electric field was parallel with the projection of cochlear modiolus on the tympanic membrane, group B: the electric field was perpendicular to modiolus projection toward to the mastoid, group C: the electric field was perpendicular to modiolus projection toward to the zygomatic process. A series of optimized recording and stimulation parameters were uesed to reduce the electrical artifact.All the 20 ears were normal in the ABR testing, and EABR waves were stable and well-differentiated in the EABR tests out of cochlea. But EABR waves of group A were more stable and differentiated than those of group B and C. In group A, the threshold of EABR was (0.54 ± 0.11) mA, and latency of wave III was (1.71 ± 0.05) ms when the stimulus intensity was 0.8 mA. In group B, the threshold of EABR was (0.62 ± 0.12) mA, and latency of wave III was (1.77 ± 0.03) ms. In group C, the threshold of EABR was (0.70 ± 0.14) mA, and latency of wave III was (1.86 ± 0.04)ms. The threshold of EABR and latency of wave III were significantly different among the three groups by statistic analysis.EABR waves were stable and well-differentiated in the EABR tests out of cochlea. The EABR waves were recorded more stably and differentiated when the stimulating electrode and recording electrode were paralleled with the projection of modiolus on the tympanic membrane.