Neural modelling of the encoding of fast frequency modulation

Frequency modulation (FM) is a basic constituent of vocalisation in many animals as well humans. In human speech, short rising and falling FM-sweeps around 50 ms duration, called formant transitions, characterise individual speech sounds. There are two representations FM the ascending auditory pathway: spectral representation, holding instantaneous frequency stimuli; sweep consisting neurons that respond selectively to direction. To-date computational models use feedforward mechanisms explain encoding. However, from neuroanatomy we know there massive feedback projections pathway. Here, found classical FM-sweep perceptual effect, pitch shift, cannot be explained by standard processing models. We hypothesised shift caused predictive mechanism. To test this hypothesis, developed novel model encoding incorporating interaction between representation. The was designed encode sweeps rate, shape transitions. It fully accounted for experimental data acquired experiment with participants previously published results. also new class stimuli second further validate model. Combined, our results indicate direction neural plays an important role FM. brain, mechanism likely occur at early stages hierarchy.