Loudness Processing of Time-Varying Sounds: Recent advances in psychophysics and challenges for future research

Current loudness models and indicators of time-varying sounds - such as the 95th percentile of the loudness distribution N5 or the A-weighted equivalent sound pressure level LAeq - all compute global loudness from short-term loudness time-series by leaving the temporal dimension out. Putting it differently, these models all rely on the assumption that the global loudness of a time-varying sound does not depend on the shape of its loudness pattern, i.e. that it can be predicted by its loudness distribution only. This assumption has been challenged and invalidated by a number of recent studies investigating time-varying sounds with both flat (e.g., fluctuating stimuli) and non-flat (e.g., rising- or falling-intensity stimuli) overall intensity profiles. We will intend to review the main outcomes of these studies and illustrate the implication of various high-level processes (cognitive, memory) in overall loudness evaluations of time-varying sounds, not yet considered in current loudness models. Then, we will discuss how the association of "traditional" (e.g. Magnitude Estimation procedures) and "molecular" (i.e. reverse-correlation methods) psychophysics could help characterize and identify the whole underlying perceptual machinery.