Predictions of the short-window model approach (DynBU_fast; green dashed lines with squares) with a sluggishness integration of short-time BMLDs are shown along with predictions using a 225 ms window for BMLD extraction (DynBU_slow; orange dashed lines with diamonds) and measured results. Data are given as a function of cutting time of the room impulse response. The left column (panels a, c and e) shows predictions for an absorption coefficient of α = 0.1 and the right column (panels b, d and f) for α = 0.5. The first row (panels a and b) shows the prediction for sound source and noise being co-located in the front of the listener (N₀S₀), the second row (panels c and d) for a sound source at 60° (N₀S₆₀). The third row (panels e and f) shows predictions for the second experiment (N₀S₀ with only late reflections). The experimentally measured binaural unmasking (solid lines) are replotted from Figure 2, panels a–d, and Figure 4, panels e and f, for comparison.