Processes of segmentation and linearization in describing events

Abstract

Natural language production is widely considered from an information processing point of view: In producing an utterance, mental representations of perceived or conceived states of affairs are first transferred to an intermediate level of propositional representations and are subsequently transformed into grammatical structures. The latter serve as input structures for phonological encoding, which results in the acoustic realization of utterances. This description of the production process follows Levelt’s model, which is outlined in Levelt (1989). Within this model, which compiles a multitude of earlier research (e.g. Fromkin, 1971; Garrett, 1980; Butterworth, 1980; Shattuck-Hufnagel, 1986), three major components interact, that are each responsible for well-defined phases within the global process of speech production. The conceptualizer draws on different knowledge sources that provide representations in multimodal formats and generates propositional conceptual structures, so-called preverbal messages, which represent the content the speaker intends to verbalize. These non-linguistic preverbal messages are transferred to the formulator Within the formulator, language specific encoding takes place; the end results of this process serve as input for the articulator. The latter is the component that carries out the phonological processing and the acoustic realization of an utterance (cf. also the description of the language production model in Glatz, Klabunde & Porzel, this volume).

The research reported in this paper has been supported by the Deutsche Forschungsgemeinschaft (DFG) in the project ‘Konzeptualisierungsprozesse in der Sprachproduktion — Produktion von Ereignisbeschreibungen’ (KONZEPTE/ Conceptualization Processes in Language Production: The Production of Event Descriptions) (Ha 1237/10). We are indebted to Carola Eschenbach, Barbara Kaup and Mark Siebel for many fruitful and insightful discussions and their helpful comment on earlier versions of this paper. Furthermore, we want to thank Mark Erichsen, Bärbel Rieckmann, and Anni-Yasmin Turhan for designing and implementing the software, as well as for their support in preparing and carrying out the empirical study.