TABLE I.

Rationale for the different stimulus conditions included in this study. Collectively, the different listening conditions represent a diversity of scene acoustics, including important examples in our environment and clinical applications. Moreover, they span maskers with different modulation statistics (Jørgensen et al., 2013; Rosen et al., 2013) and stimuli with intact and degraded TFS, which allowed us to rigorously test our hypotheses. Note that the SNR levels were chosen to span the full range of intelligibility without floor or ceiling effects.

No.	Stimulus condition	Rationale for inclusion in study
1–3	SiSSN at SNRs of −2, −5, and −8 dB	Widely used in the literature; used for calibration of prediction model
4 and 5	SiB at SNRs of 4 and −2 dB	Simulates ecologically relevant cocktail-party listening; has different masker modulation statistics from SiSSN
6	SiB at 6 dB SNR subjected to reverberation (T60 = 2.4 s)	Reverberation is ubiquitous in everyday listening environments (e.g., rooms and stairwells); linearly distorts temporal information
7	SiB at 4 dB SNR subjected to 64-channel envelope vocoding	Used to investigate the role of TFS in target-speech coding and intelligibility
8	SiB at −6 dB SNR subjected to 64-channel ITFS	ITFS is a precursor to deep-learning-based denoising algorithms that are increasingly used in many audio processing applications, including hearing aids (Wang and Chen, 2018); nonlinear distortion