One of the biggest complaints by cochlear implant (CI) users is their difficulty to understand speech in noisy situations (Kochkin, 2005). Because speech-in-noise understanding requires multiple sensory and cognitive neural processes, there can be various causes for the difficulties experienced by CI users, which poses a significant barrier to clinical assessment and treatment.

Prior works in normal hearing (NH) listeners have identified several central processes that constitute successful speech-in-noise perception by unmasking speech from background noise. Those processes include: (1) accurate encoding of acoustic features of sound sources such as pitch (Lad et al., 2020), (2) abstraction of the target from crowded auditory scenes (Holmes and Griffiths, 2019), and (3) neural enhancement of target speech and inhibition of background noise (Choi et al., 2014; Kim et al., 2021). The above processes may cascade, meaning that the failure of acoustic feature encoding may degrade downward processes of target speech abstraction, neural enhancement of target speech and inhibition of background noise. Indeed, CI listening, which lacks the delivery of low-frequency for pitch perception, often exhibits poor performance in auditory stream segregation (Oxenham, 2008).

In the recent decade, hearing remediation through CIs has moved toward combining electric and acoustic stimulation (EAS) (Gantz and Turner, 2004; von Ilberg et al., 1999). As expected, EAS listeners with residual low-frequency acoustic hearing demonstrate improved encoding of acoustic features such as pitch (Brockmeier et al., 2010; Gantz et al., 2005; McDermott et al., 2009; Woodson et al., 2010). It has been also demonstrated that CI users with EAS perform better at speech-in-noise (SiN) tasks than electric-only (herein referred to as the “E-only”) CI users (Gantz et al., 2006; Gantz and Turner, 2004; Incerti et al., 2013; Rader et al., 2013; Turner et al., 2008). In the studies that compared EAS listening to E-only listening in the same subjects, the EAS condition exhibited significantly lower speech reception thresholds than the E-only condition (Imsiecke et al., 2020a).

However, the neural mechanism underlying improved SiN performance in EAS listeners has been unclear. Specifically, it is unclear whether better SiN performance in EAS listeners is due to the better central processes for unmasking. This study aimed to directly compare neural processes of speech unmasking between EAS and E-only listeners.

As described above, the fidelity of acoustic feature encoding may affect the separation of target speech from background noise. Supporting this idea, previous electro-physiological studies in NH listeners have correlated the fidelity of supra-threshold acoustic cue coding to SiN understanding (Anderson and Kraus, 2010; Anderson and Parbery-Clark, 2013; Hornickel et al., 2009; Song et al., 2011). Also focusing on the contribution of acoustic feature encoding to SiN understanding, previous CI studies showed that the cortical acoustical change complex (McGuire et al., 2021) and N1-P2 cortical auditory event-related potential (Berger et al., 2021) predict SiN performance.

Endogenous selective attention also contributes to the speech unmasking process. It has been reported that reaction time during a selective attention task predicts SiN performance in NH listeners (Strait and Kraus, 2011). Similarly, previous CI studies suggest that the strength of attentional modulation on cortical evoked responses to competing sounds predicts SiN ability (Lee et al., 2021; Nogueira and Dolhopiatenko, 2022).

A recent study in NH listeners suggested an electroencephalographic (EEG) index of overall speech unmasking function that combines these bottom-up and top-down neural processing, referred to as internal SNR (Kim et al., 2021). Internal SNR is the amplitude ratio of cortical evoked responses to speech (i.e., a single word) and background noise. Higher internal SNR indicates better neural unmasking of target speech from background noise. The internal SNR showed a significant positive correlation with SiN performance in NH listeners. Interestingly, better NH performers in the SiN task exhibited greater internal SNR due to their weaker cortical evoked response to background noise, not due to their stronger evoked response to the target speech, indicating that inhibiting the neural representation of noise is a driving factor of better speech unmasking.

This study aimed to compare internal SNR between EAS and E-only CI users. Findings from this study may reveal how cortical process for unmasking speech from background noise is different between EAS and E-only listening, which has not been reported by previous studies.