Lecturer's Précis - Garrett (1990)
Copyright Notice: This material was written and published in Wales by Derek J. Smith (Chartered Engineer). It forms part of a multifile e-learning resource, and subject only to acknowledging Derek J. Smith's rights under international copyright law to be identified as author may be freely downloaded and printed off in single complete copies solely for the purposes of private study and/or review. Commercial exploitation rights are reserved. The remote hyperlinks have been selected for the academic appropriacy of their contents; they were free of offensive and litigious content when selected, and will be periodically checked to have remained so. Copyright © 2010, High Tower Consultants Limited.
First published online 14:30 BST 17th October 2002, Copyright Derek J. Smith (Chartered Engineer). This version [HT.11 - reattach diagram] dated 09:00 8th December 2010
Although this paper is reasonably self-contained, it is best read as a subordinate file to our e-paper on "Speech Errors and Speech Production Models". An earlier version of this material appeared in Smith (1997; Chapter 6). It is repeated here with a colour-coded graphic and supported with hyperlinks.
Garrett's (1990) Model of Sentence Production
The University of Arizona's Merrill F. Garrett based his model of speech production on the sort of evidence provided by speech error studies. He believes that a speaker's communicative intention - the message - is input to a post-semantic speech production system. Conventionally enough, he describes the first stage of this process as obtaining some sort of word meaning record which does not yet include information as to the final word form. He then concentrates upon the operations which convert such messages into instructions capable (a) of initiating lexical retrieval, and (b) of guiding the articulatory system. Errors can occur at either stage: at the first stage they will be meaning-related, whilst at the second stage they will be form-related.
Like the transcoding modellers, Garrett supports his theorising with a dataflow diagram (DFD). There were several early attempts at such models (eg. Garrett, 1975, 1980, 1982, 1984), but a more up-to-date version is provided by Garrett (1990), in which he settles on a five-level model of processing. There is one semantic level, namely the message level, and four post-semantic levels, namely the functional, positional, phonetic, and motor levels. These levels are described in the caption box below:
Garrett's Speech Production Model: This model is a DFD with additional annotation, and Garrett's explicit assumption is that "a speaker's particular communicative intention [acts] as the input to the production system; we will refer to that input as a message and the processes associated with it as the Message level" (Garrett, 1990, p158; italics original). Here are the processing levels identified:
THE SEMANTIC LEVEL (THE COMPREHENSION SYSTEM)
Message Level (M): This is the top level of sentence construction, which Garrett groups together as "Inferential Processes" [topmost mauve box]. This is the level at which the mind somehow makes up its mind what it wants to say. It is the level of meaning, and whenever an act of willed and meaningful communication is decided upon, a suitably coded message, M, is passed downwards for further action.
THE POST-SEMANTIC LEVELS (THE PRODUCTION SYSTEM)
Functional Level (F): This is where the "phonologically uninterpreted" syntactic relationship between words is determined. There are two sub-processes, lexical selection and functional structuring, both colour-coded mid-blue. This level processes more than one, but probably only two, phrases at a time (because word exchanges occur between adjacent clauses, but rarely involve elements of non-adjacent clauses), and when processing has been completed a functional level representation, F, is passed down to the next level for further action.
Positional Level (P): This is where the surface form of the sentence starts to appear, and matters of morphology, syntax, and word order are sorted out. There are three sub-processes, word form retrieval, positional planning, and node information, all colour-coded light-blue. Garrett justifies this level being separate from the functional level by referring to the phenomenon of accommodation in speech errors. Thus if the message "a language acquisition device" were to be articulated as "an anguage lacquisition device", the fact that the morpheme "an" has replaced "a" correctly (but stupidly) indicates that the allocation of "a" and "an" is normally made at a later stage in production than word-building (morphology) or phrase-building (syntax). The positional level is in fact the first of two levels at which the final sound structure of the sentence starts to be determined. This level deals with surface syntax, prosody, and abstract segmental and lexical structure. It is seen as processing only one phrase at a time (which has as-yet-unexplored implications for the feedback structure used), and when processing has been completed a positional level representation, P, is passed down to the next level for further action.
Phonetic Level: This is the second sound structure level, and has been colour-coded lime. It deals with detailed phonetic properties.
Motor Level: This is the level of initiating actual muscular activity, and has been colour-coded light lime. It is the level where the functional motor systems communicate with the anatomical motor systems.
Stripped of its annotation, the model closely follows Lordat's (1943) "boxology" (Lecours, Nespoulos, and Pioger (1987). Garrett's M-level is equivalent to Lordat's Stage 1, and Garrett's F, P, and Motor levels are Lordat's Stages 2 to 4. Note the two major instances of parallel processing. The first is where the message M simultaneously initiates both word selection processes (upper left fork) and sentence structuring processes (upper right fork). The second is where the functional level representation F simultaneously initiates word form retrieval processes (lower left fork) and clause structuring processes (lower right fork).
Developed from a black and white original in Garrett (1990:164), but with the Motor level, the italicised notes, and the colour coding added. This version Copyright © 2002, Derek J. Smith.
That is the end of our commentary. Now do the following exercise .....
Exercise to Consolidate
Garrett (1990) suggests that "the production system uses the comprehension system to monitor its own products. We do listen to what we say ourselves, and we do this not only to adjust the form of our utterance more closely to our communicative objectives but also to monitor for error in the operation of the systems that generate spoken form." (p166). However, the diagram includes none of the feedback pathways by which this might take place. So .....
1. Take a spare copy of the Garrett model, and mark out all the feedback loops and servomechanisms you can identify.
2. Mark the processing carried out in parallel.
3. Dot in all possible additional feedforward and feedback loops (ie. those Garrett has omitted).
Garrett, M.F. (1975). The analysis of sentence production. In Bower, G.H. (Ed.), The Psychology of Learning and Motivation (Volume 9). New York: Academic Press.
Garrett, M.F. (1980). Levels of processing in sentence production. In Butterworth, B. (Ed.), Language Production (Volume 1). Orlando, FL: Academic Press.
Garrett, M.F. (1982). Production of speech: Observations from normal and pathological language use. In Ellis, A.W. (Ed.), Normality and Pathology in Cognitive Functions. London: Academic Press.
Garrett, M.F. (1984). The organisation of processing structure for language production: Applications to aphasic speech. In Caplan, D., Lecours, A.R., and Smith, A. (Eds.), Biological Perspectives on Language. Cambridge, MA: MIT Press.
Garrett, M.F. (1990) Sentence processing. In Osherson, D.N. and Lasnik, H. (Eds.). An Invitation to Cognitive Science (Volume 1): Language. Cambridge MA: MIT Press.
Lecours, A.R., Nespoulos, J.L., and Pioger, D. (1987). Jacques Lordat, or the birth of cognitive neuropsychology. In Keller, E. and Gopnik, M. (Eds.), Motor and Sensory Processes in Language. Hillsdale, NJ: Erlbaum.
Smith, D.J. (1997). Human Information Processing. Cardiff: UWIC. [ISBN: 1900666081]