An integrated model of speech melody – Transmitting multiple communicative meanings by controlling articulatorily operable parameters

Friday 20th February: Yi Xu
An integrated model of speech melody – Transmitting multiple communicative meanings by controlling articulatorily operable parameters

Abstract

When trying to understand phonemes such as consonants, vowels and lexical tones, one is constantly struck by the amount of variability in both their acoustics and articulation. When trying to understand information transmission through speech, one is often struck by the fact that so many different communicative meanings are transmitted virtually simultaneously through acoustic signals.

While past research has gone a long way toward identifying the individual pieces of both puzzles, some critical links are missing for joining the two into a larger picture. In this talk, I will argue that a comprehensive model that can provide the missing links has to satisfy at least three critical requirements. First, it needs to make a clear separation between the meaning-bearing components, which are functionally defined, and the encoding primitives, which are defined in forms and detached from meanings. Second, it needs to specify an uninterrupted link between articulatory mechanisms of speech production and the functional components of communicative meanings. Third, it needs to specify mechanisms for concurrent transmission of multiple (as opposed to just one or two)
communicative functions. To illustrate how these requirements can be met, I will present the Parallel Encoding and Target Approximation (PENTA) model.
The model defines specific ways through which communicative functions are both articulatorily implementable and perceptually decomposable. By so doing, it treats much of the observed variability as sources of individual communicative meanings rather than as either random noise or extraneous paralinguistic factors non-central to speech communication. The data in support the current version of the PENTA model are mostly from research on tone and intonation. But I will argue that it is just as applicable to other aspects of speech.