Bimodal hearing—combining a cochlear implant (CI) in one ear and a hearing aid (HA) in the other—can enhance speech perception by providing complementary acoustic and electric cues. However, benefits vary widely across listeners, partly due to mismatches in how spectral information is encoded across ears. This talk presents three studies examining how vowel formant cues are integrated in bimodal listening, using two simulated approaches (Study 1 and Study 2) and one real‑listener approach (Study 3).
Study 1 tested whether vowel recognition improves when the spectral distance between a vowel’s F1 and F2 is small rather than large. Normal‑hearing listeners identified synthetic vowels under EAS, bimodal, CI‑alone, and HA‑alone conditions, with F1 delivered to the acoustic ear and F2 to the electric ear. Vowels with closely spaced formants produced stronger spectral integration, whereas widely separated formants degraded bimodal performance and introduced interference when dominant cues were split across ears. Within‑ear (EAS) integration consistently outperformed across‑ear (bimodal) integration.
Study 2 examined a different dimension of spectral integration: interaural F1 mismatch. Using simulated bimodal stimulation with normal-hearing listeners, F1 was presented to the acoustic ear, while the CI ear received both F1 and F2. F1 in the CI ear was systematically shifted relative to the HA ear. Vowel recognition declined as the interaural F1 mismatch increased, indicating that matched F1 across ears supports effective integration, whereas mismatched F1 forces listeners to rely on one ear and reduces bimodal benefit. This study highlights the sensitivity of vowel perception to low‑frequency spectral alignment across modalities.
Study 3 tested real bimodal listeners using a reverse‑correlation (RC) algorithm to reconstruct the perceptually integrated vowel spectrum. Listeners compared an original vowel to many randomly generated variants, judged each pair as “similar” or “different,” and subtracting the average of different trials from similar trials yielded each listener’s internal integrated spectrum. Reconstructed spectra differed markedly from the original vowels and showed substantial individual variability, suggesting a direct way to characterize listener‑specific integration patterns for individualized fitting.
Together, these studies reveal how spectral spacing, interaural alignment, and individual auditory characteristics shape bimodal spectral integration and point toward personalized fitting strategies based on each listener’s integrated spectral profile.