Item

Abstract

Frequency modulation (FM) is a basic constituent of vocalisation. Formant transitions in speech are characterised by short rising and falling FM-sweeps in the kilohertz frequency range. These sounds elicit a pitch percept that deviates from their average frequency. This study uses this perceptual effect, termed here the sweep pitch shift, to inform a model characterising the neural encoding of FM. First, a reexamination of the classical effect, consisting of two perceptual experiments, provides a quantitative characterisation of the dependence of the sweep pitch shift with the properties of the sweeps. Next, simulations carried on the new experimental data show that classical temporal and spectral models of pitch processing cannot explain the pitch shift. Conversely, a modified spectral model considering a predictive interaction between frequency and FM encoding fully reproduces our and previous experimental data. The model introduces a feedback mechanism that modulates the neurons that are expected to respond to future portions of the sweeps, accelarating their onset response. Combined, the experimental and modelling results suggest that predictive feedback modulation plays an important role in the neural encoding of FM even at early stages of the processing hierarchy.