Incremental Sentence Production, Self-Correction and Coordination

Since Garrett’s (1975, 1980) seminal work on speech error phenomena, it has become customary to distinguish four levels of representation within the sentence production process: a message level, a functional level, a positional level, and a phonetic level (see also Bock, this volume). Garrett’s model has been further elaborated and modified by Kempen (Kempen & Hoenkamp, in press; Van Wijk & Kempen, in press) who proposes the global sentence production model depicted in Figure 1. The four modules listed have the following functions: 1. The conceptual module forms a conceptual (semantic) representation of the message which the speaker wishes to communicate. The nature of the semantic structures output by this component need not concern us here. 2. The lexico-syntactic module constructs an ordered tree structure consisting of constituents and their functional relations. The terminal nodes of syntactic trees (both content and function words) are instances of abstract (not phonologically specified) lexical items called lemmas which are retrieved from the lexicon. While Garrett assigns the tasks of inserting function words and computing word order to a later module (the positional stage), Kempen assigns them to this one. 3. The morpho phonological module computes the word form of all lemmas by retrieving their phonological specifications (lexemes) from the lexicon and making various morphological and phonological adjustments. 4. The articulatory module produces a phonetic specification which is used to control the articulatory apparatus.