Abstract
Loudness increases with duration, an effect commonly referred to as temporal integration of loudness and characterized by time constants of leaky integrators. The present study compares model predictions with and without compression with experimental data from a round robin test. It shows that a simple leaky integrator model approach is unable to predict the slope of the temporal integration function for loudness when the compressive characteristics of the auditory system are taken into account. A deviation is still observed when a more sophisticated loudness model is used. A good prediction is achieved when the single low-pass filter stage of the loudness model is replaced with a stage consisting of two parallel low-pass filters with different time constants.