Objective: Cochlear microphonic responses (CMs) were measured in a rabbit model in the intact ear (CM1) and in ears with a partially implantable piezoelectric middle ear implant (P-MEI) (CM2) to investigate the characteristics of speech transmission of the P-MEI in vivo. Design: The spectra of pure tones, voices, and elicited CM1 and CM2 obtained from the round window before and after the implantation of a P-MEI device were calculated by using fast Fourier transform. Frequency response functions of CM1 and CM2 were used to demonstrate the functional similarity between the implanted P-MEI and the normal ossicular chain. The coherence functions between the voices and CM1 and between CM1 and CM2 were evaluated to characterize speech transmission of the P-MEI in vivo. Ten rabbit ears were used in this study. Pure tones, six Chinese vowels, and six Chinese characters were the acoustic stimuli. The CMs elicited by a list of bisyllabic words were tape-recorded and then recognized by subjects with normal hearing. Results: Using pure tones at the same intensity of 90 dB SPL, frequency response functions of the CMs between the two states (the intact ear with normal hearing and ears with a P-MEI device at the medium volume) were calculated showing great resemblance in shape. Compared with that at 1 kHZ, gain factors were 10 and 20 dB, respectively, at higher frequencies. The correlation and spectral analyses of the vocalizations, CM1 and CM2, demonstrated that the harmonics of CM1 were approximately identical as those of the voices between 0.5 and 5.0 kHz with coherence functions of about 0.7 similar to 1 at the formants' frequencies, whereas the harmonics of CM2 between 0.5 and 2.5 kHz were enhanced with the coherences near to unity at the formants' frequencies, and others <0.5 kHz and >2.5 kHz were attenuated. The recognition score of the CMs elicited by a list of bisyllabic words was >90% using subjects with normal hearing. Conclusions: Data from this study suggest that cochlear microphonic potentials can be used as an important tool to evaluate objectively whether the implantation of a P-MEI device is successful and whether the quality in speech transmission of the P-MEI is satisfactory. Thus, the method would be of significance to clinically ascertain the performance of the device in vivo.