The Roles of Selective Attention and Asymmetric Experience in Bilateral Speech Interference for Single-Sided Deafness Cochlear Implant and Vocoder Listeners

Objectives: For many (especially older) single-sided-deafness (SSD) cochlear-implant (CI) users (one normal hearing and one CI ear), masking speech in the acoustic ear can interfere with CI-ear speech recognition. This study examined two possible explanations for this “bilateral speech interference.” First, it might reflect a general (i.e., not specific to spatial hearing or CI use) age-related “selective-attention” deficit, with some listeners having difficulty attending to target speech while ignoring an interferer. Second, it could be specific to asymmetric-hearing experience, reflecting maladaptive plasticity with the better ear becoming favored over time. Design: Twenty-eight listeners with bilaterally normal or near-normal hearing (NH) through 4 kHz completed a series of speech-on-speech masking tasks. Vocoder simulations of SSD-CI listening (four- or eight-channel noise-vocoded speech in the right ear, unprocessed speech in the left) tested whether acutely simulated asymmetric hearing would produce interference comparable to that previously observed for 13 SSD-CI listeners. Both groups had a wide age range (NH: 20 to 84 years; SSD-CI: 36 to 74 years) and were therefore expected to exhibit a wide range of selective-attention ability. The primary set of conditions measured bilateral speech interference. Target coordinate-response-measure sentences mixed with a masker of similar fundamental frequency (F0) were presented to the right (vocoded) ear at target-to-masker ratios of 0, 4, 8, or 16 dB. Silence or a copy of the masker was presented to the left (unprocessed) ear. Bilateral speech interference—the performance decrease from adding the masker copy to the left ear—was compared with previous SSD-CI results. NH listeners also completed two additional sets of conditions. The first set measured the F0-difference benefit for unprocessed monaural speech-on-speech masking. This is a likely indicator of non-spatial selective-attention ability, based on previous findings that older adults benefit less than younger adults from target-masker F0 differences. The second set measured contralateral-unmasking benefit. Target and masking speech were presented to the unprocessed ear and the benefit from presenting a copy of the masking speech to the vocoded ear was measured. A linear-mixed model analysis examined relationships between NH bilateral speech interference and age, monaural speech-on-speech masking (to estimate non-spatial selective attention), and contralateral unmasking. An additional analysis compared NH-Vocoder to SSD-CI interference. Results: The strongest predictor of NH-vocoder interference was performance in the monaural different-F0 speech-on-speech masking condition ( p = 0.0024). Neither similar-F0 speech-on-speech masking performance, nor age, nor contralateral unmasking accounted for significant additional variance ( p = 0.11 to 0.69). Mean SSD-CI interference magnitude was comparable to the four-channel NH-vocoder condition ( p = 0.75) but larger than the eight-channel condition ( p < 0.001). Conclusions: The association between bilateral speech interference and monaural speech recognition with different-F0 interferers for NH listeners suggests that (possibly age-related) non-spatial selective-attention ability might also explain SSD-CI interference variability. Regardless of hearing status, some people might have difficulty attending to one sound while ignoring others, with asymmetrically distorted inputs exacerbating this problem. Comparable SSD-CI and NH-Vocoder interference challenges the idea that SSD-CI interference reflects long-term maladaptive changes from asymmetric hearing. Future work should explore selective-attention measures as predictors of SSD-CI performance in competing-talker environments.