FIG. 9.

(Color online) Results support an integrative conceptual model of speech intelligibility. Taken together, our results support this integrative conceptual model of speech intelligibility in that they clarify what internal representation is predictive of speech intelligibility and how that representation is related to the acoustics of the auditory scene and cognitive variables. Our results show that the strength of the net envelope (ENV) coding of target speech relative to other interfering sounds in the central auditory system predicts intelligibility in a variety of real-world listening conditions (arrow A). The modulation frequencies that contribute to these EEG-based intelligibility predictions depend strongly on the envelope spectrum of the masker and the scene acoustics. TFS cues (arrow B) also affect how well neural responses in the central auditory system encode the envelope of target speech, likely by aiding in source segregation (Darwin, 1997; Micheyl and Oxenham, 2010; Oxenham and Simonson, 2009). Selective attention can then operate effectively on the distinct representations of segregated target and masker objects (arrow C) to boost the neural representation of the target relative to the masker (Ding and Simon, 2012; O'Sullivan et al., 2015; Viswanathan et al., 2019). Taken together, our results support the theory that scene analysis and attentive selection of target speech are influenced by both modulation masking and TFS, consistent with the broader temporal coherence theory.