Our methods in speech production and processing
In order to understand individual variability we are collecting different types of data related to our participants. In addition to demographic information, we are collecting the following:
Empathy Quotient and Autism Quotient, since both empathy and autistic traits are said to play a role in the use of intonation
Auditory abilities, particularly frequency discrimination acuity
Musicality, defined as a combination of tonal memory, affinity for music, and musical training.
To understand contextual variability, we are collecting a large corpus of both scripted and spontaneous data. The data will be annotated using AM and RPT, and modelled using Functional Principal Component Analysis.
Naïve Bayes classifiers trained on scripted data will be used to determine whether disputed categories (e.g., types of English final rises) can be discriminated based on form. Finally, unsupervised clustering will be used on the same data to examine how closely they correspond to these putative tonal categories.
The speech corpora will be pragmatically annotated and analysed. Behavioural experiments, eyetracking and neurophysiological studies will be used to test (a) compositionality in intonational meaning, (b) the perceptual relevance of the phonetic findings, particularly with respect to gradience and cue-trading, and (c) the role of intonation in communication across linguistic varieties.
SPRINT’s final phase will involve phonological analysis of the phenomena studied and the development of a model of intonational meaning. The analyses will be developed taking into consideration both the results of the experiments and modelling, and principles of phonology and pragmatics.