Categories and gradience in intonation

The Autosegmental-Metrical framework (AM) assumes that a distinction needs to be made between linguistic phonological information (categorical) and paralinguistic phonetic information (gradient) in intonation. However, empirical evidence supporting this assumption has proved to be elusive so far. In this study we analysed whether the theoretical distinction is reflected in perceptual biases and neural activation in the brain. The results of a combined behavioural and neuroimaging study demonstrate that intonational function indeed activates different but overlapping neural networks with more widespread activation for categorical phonological stimuli, especially in middle temporal gyrus bilaterally and left supramarginal and inferior parietal areas. In contrast, for paralinguistic gradient stimuli activation is restricted to right inferior frontal gyrus. These neural differences mirror differences in response times in a listening experiment testing categorical perception for the same stimuli. These findings support a theoretical model of intonation, such as AM, in which linguistic and paralinguistic information are distinguished.