Interaural Place-Mismatch Estimation With Two-Formant Vowels in Unilateral Cochlear- Implant Users

Background: For patients with one cochlear implant (CI) and residual hearing in the opposite ear, a default frequency-to-electrode map is typically used despite large individual differences in electrode-array insertion depth. This non-individualized fitting rationale might partly explain the variability in long-term speech-reception benefit among CI users. Knowledge about the electrode-array location is thus crucial for adequate fitting. Although electrode location can theoretically be determined from CT scans, these are often unavailable in audiological practice. Moreover, existing behavioral procedures such as interaural pitch-matching are rather tedious and time-consuming. Here, an alternative method using two-formant vowels was developed and tested. Methods Eight normal-hearing (NH) listeners were presented synthesized two-formant vowels embedded between consonants /t/ and /k/, with first-formant frequencies (F1) at 250 and 400 Hz and second-formant frequencies (F2) between 600 and 2200 Hz. F1 was presented unaltered to the left ear, while F2 was presented to the right ear via a vocoder system simulating 3 different CI insertion depths. In each condition, the listeners indicated in a forced-choice task which of 6 vowels they perceived for different [F1, F2] combinations. Ten CI users (5 bimodal and 5 single-sided deaf) performed the same task for F1 presented acoustically to the non-CI ear and F2 presented either acoustically to the same ear or electrically to the CI ear. Results After some training, all NH listeners were able to fuse the unaltered F1 and vocoded F2 into a single vowel percept, and vowel distributions could be reliably derived in 7 listeners. Vocoder simulations of reduced CI insertion depth led to clear vowel-distribution shifts in these listeners. However, these shifts were overall smaller than their theoretical value, with high across-subject variability. Vowel distributions could be derived for all CI users in the monaural acoustic condition, indicating an ability to perform the task reliably. Despite this, large individual differences were observed for dichotic bimodal stimulation, with listeners showing either basal or apical shifts, or generally-poor vowel discrimination. Conclusion The two-formant-vowel method is a fast and clinic-friendly candidate to derive interaural place mismatches from a simple vowel-recognition task. However, it remains unclear whether the measured “vowel spaces” in CI users are directly related to insertion depth, and whether they are influenced by the ability to fuse acoustic and electric stimuli or habituation to the CI. The comparison of the present results to CT-scan and speech-intelligibility data in the same listeners will shed light on the validity of the proposed method.