Binaural masking release is modeled as within-band noise floor reduction

Evidence

Levitt and Rabiner ( 1967 ) predicted the spatial improvement in speech intelligibility using this hypothesis. The underlying assumption of their model is that a BMLD effect acts within each frequency band, leading to a within-band detection advantage. This was modeled as a reduction of the level of the noise floor equal in size to the BMLD for that particular frequency, resulting in a more advantageous signal-to-noise ratio (SNR). This would make previously undetectable speech elements audible and enable them to contribute to speech intelligibility. — Dijk, Pim van, Physiology, Psychoacoustics and…, 1622