Introduction

Preview

In this chapter we shall see how far the naturally displayed evidence from language production (which we reviewed in ch. 3) can be used to establish components of a model of language production.

In our exposition, we shall start with the model proposed by Garrett (1982): this is a convenient starting point for us because it represents the culmination (thus far) of the error-based model discussed in Garrett's earlier work, but also attempts to pull together insights from the work of Fromkin, also based on error data, and that of Goldman-Eisler and her associates (Goldman- Eisler 1968; Beattie 1980; Butterworth 1980b) on the evidence of hesitation phenomena. In section 7.2 we start looking more closely at the topmost, or ‘message-structure’ level. In this, we make contact with the meaningrepresentation issues that we arrived at in the last chapter, in considering comprehension processing. Much of what we have to say about possible mental representation here derives from the work of Johnson-Laird (1983). We return to the matter of lexical access in production, briefly touched on in chapter 5, in section 7.3, and consider the form in which word meanings may be represented in the mental lexicon. Garrett's model recognises two levels of sentence structure, a deeper as well as a more superficial one, and their organising characteristics are investigated in the next section (7.4). Finally, we consider the serial nature of Garrett's model and examine some arguments (Dell and Reich 1981; Stemberger 1985) relating to the possible interactions between levels in such a model (section 7.5) which would lead to a parallel interpretation.