Figure 2.

Construction of the speech/noise stimuli that were used in this study. Original speech snippets and complementary noise (based on the instantaneous envelope phase of the individual snippet and the average original spectral content at the respective phase; see Materials and Methods for details) were added, yielding speech snippets with conserved high-level features (fluctuating in the frequency range of ∼2–8 Hz) but comparable spectral content across envelope phases. Parts processed in the real domain are specified in blue and as a signal in time. Parts processed in the wavelet domain are specified in red and are represented as a time–frequency signal.